The Basic Principles To Programming

IIT Delhi
S. Arun-Kumar, CSE
Introduction to Title Page
Computer Science JJ
Contents
II
S. Arun-Kumar J I
sak@cse.iitd.ernet.in Page 1 of 709
Department of Computer Science and Engineering Go Back
I. I. T. Delhi, Hauz Khas, New Delhi 110 016.

Full Screen
March 19, 2007 Close
Quit
Contents
1 Computing: The Functional Way 3
1.1 Introduction to Computing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Our Computing Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.3 Primitives: Integer & Real . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 IIT Delhi
1.4 Example: Fibonacci . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
1.5 Primitives: Booleans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
S. Arun-Kumar, CSE
2 Algorithms: Design & Refinement 114

2.1 Technical Completeness & Algorithms . . . . . . . . . . . . . . . . . . . . . . . . 114 Title Page
2.2 Algorithm Refinement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
2.3 Variations: Algorithms & Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
2.4 Names, Scopes & Recursion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208 Contents
3 Introducing Reals 242

3.1 Floating Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242 JJ II
3.2 Root Finding, Composition and Recursion . . . . . . . . . . . . . . . . . . . . . . 264
J I
4 Correctness, Termination & Complexity 294
4.1 Termination and Space Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . 294
4.2 Efficiency Measures and Speed-ups . . . . . . . . . . . . . . . . . . . . . . . . . 333 Page 2 of 709
4.3 Invariance & Correctness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
5 Compound Data 390 Go Back
5.1 Tuples, Lists & the Generation of Primes . . . . . . . . . . . . . . . . . . . . . . . 390

5.2 Compound Data & Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
Full Screen
5.3 Compound Data & List Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . 457
6 Higher Order Functions & Structured Data 486 Close

6.1 Higher Order Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 486
6.2 Structured Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513
Quit
6.3 User Defined Structured Data Types . . . . . . . . . . . . . . . . . . . . . . . . . 547
7 Imperative Programming: An Introduction 570

7.1 Introducing a Memory Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 570
7.2 Imperative Programming: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 599
7.3 Arrays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 626 IIT Delhi
8 A large Example: Tautology Checking 640

8.1 Large Example: Tautology Checking . . . . . . . . . . . . . . . . . . . . . . . . . 640 S. Arun-Kumar, CSE
8.2 Tautology Checking Contd. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 665
Title Page
9 Lecture-wise Index to Slides 680
Contents
JJ II
J I
Page 3 of 709
Go Back
Full Screen
Close
Quit
1. Computing: The Functional Way
1.1. Introduction to Computing
1. Introduction
IIT Delhi
2. Computing tools
3. Ruler and Compass S. Arun-Kumar, CSE
4. Computing and Computers

Title Page
5. Primitives
6. Algorithm
Contents
7. Problem: Doubling a Square
8. Solution: Doubling a Square JJ II
9. Execution: Step 1
J I
11. Doubling a Square: Justified Page 4 of 709
12. Refinement: Square Construction

Go Back
13. Refinement 2: Perpendicular at a point
14. Solution: Perpendicular at a point Full Screen
15. Perpendicular at a point: Justification

Close
Next: Our Computing Tool
Quit
Introduction IIT Delhi
S. Arun-Kumar, CSE
• This course is about computing Title Page
• Computing as a process is nearly Contents
as fundamental as arithmetic JJ II
• Computing as a mental process J I
• Computing may be done with a va- Page 5 of 709
riety of tools which may or may not Go Back
assist the mind Full Screen
Close
Quit
Computing tools
IIT Delhi
S. Arun-Kumar, CSE
• Sticks and stones (counting) Title Page
• Paper and pencil (an aid to mental Contents
computing) JJ II
• Abacus (still used in Japan!) J I
Page 6 of 709
• Slide rules (ask a retired engineer!)

Go Back
• Ruler and compass Full Screen
Close
Quit
Ruler and Compass IIT Delhi
S. Arun-Kumar, CSE
Actually it is a computing tool!

Title Page
• Construct a length that is half of a Contents
given length JJ II
• Bisect an angle J I
• Construct a square that is twice the Page 7 of 709
area of a given square Go Back
√
• Construct 10 Full Screen
Close
Quit
Computing and IIT Delhi
Computers S. Arun-Kumar, CSE
Title Page
• Computing is much more funda- Contents
mental JJ II
• Computing may be done without a J I
computer too! Page 8 of 709
• But a Computer cannot do much Go Back
besides computing. Full Screen
Close
Quit
Primitives
• Each tool has a set of capabilities IIT Delhi
called primitive operations or primi-

S. Arun-Kumar, CSE
tives
Title Page
Ruler: Can specify lengths, lines Contents
Compass: Can define arcs and cir- JJ II
cles J I
• The primitives may be combined in Page 9 of 709
various ways to perform a computa- Go Back
tion. Full Screen
• Example Constructing a right bisec- Close
tor of a given line segment. Quit

IIT Delhi
Algorithm S. Arun-Kumar, CSE
Given a problem to be solved with a Title Page
given tool, the attempt is to evolve a Contents
combination of primitives of the tool in JJ II
a certain order to solve the problem. J I
An explicit statement of this combina- Page 10 of 709
tion along with the order is an algo- Go Back
rithm Full Screen
Close
Quit
IIT Delhi
Problem: Doubling a S. Arun-Kumar, CSE
Square Title Page
Given a square, construct another Contents
square of twice the area of the origi- JJ II
nal square. J I
A D
Page 11 of 709
Go Back
B C
Full Screen
Close
Quit
IIT Delhi
Solution: Doubling a S. Arun-Kumar, CSE
Square Title Page
Contents
Assume given a square ABCD of
JJ II
side a > 0.
J I
1. Draw the diagonal AC.
Page 12 of 709
2. Complete the square ACEF on Go Back
side AC. Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Execution: Step 1 Title Page
Contents
Draw the diagonal AC.

JJ II
A D
J I
Page 13 of 709
B C
Go Back
Full Screen
Close
Quit
IIT Delhi
Execution: Step 2 S. Arun-Kumar, CSE
Complete the square ACEF on side Title Page
AC. Contents
JJ II
J I
A D
Page 14 of 709
Go Back
B C
Full Screen
Close
Quit
IIT Delhi
Doubling a Square: S. Arun-Kumar, CSE
Justified Title Page
Contents
Assume given a square ABCD of
JJ II
side a > 0.
√ J I
1. Draw the diagonal AC. AC = 2a
Page 15 of 709
2. Complete the square ACEF on Go Back
side AC. Area of ACEF = 2a2. Full Screen
Close
Quit
Refinement: Square
Given a line segment of length b > 0
construct a square of side b. IIT Delhi
Assume given a line segment P Q of S. Arun-Kumar, CSE
length b. Title Page
1. Construct two lines l1 and l2 per- Contents
JJ II
pendicular to P Q passing through
P and Q respectively J I
Page 16 of 709
2. On the same side of P Q mark

Go Back
points R on l1 and S on l2 such that

Full Screen
P R = P Q = QS.
Close
3. Draw RS. P QSR is a square of Quit
side b
Square on Segment: 0
Assume given a line segment P Q of IIT Delhi
length b. S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 17 of 709
Go Back
P Q
Full Screen
Close
Quit
Construct two lines l1 and l2 perpen-
dicular to P Q passing through P and IIT Delhi
Q respectively S. Arun-Kumar, CSE
Title Page
l l
1 2
Contents
JJ II
J I
Page 18 of 709
Go Back
P Q Full Screen
Close
Quit
On the same side of P Q mark points IIT Delhi
R on l1 and S on l2 such that P R =
P Q = QS. S. Arun-Kumar, CSE
Title Page
l l
1 2
R S Contents
JJ II
J I
Page 19 of 709
Go Back
P Q Full Screen
Close
Quit
Draw RS. P QSR is a square of side
b IIT Delhi
S. Arun-Kumar, CSE
l l
1 2
Title Page
R S
Contents
JJ II
J I
Page 20 of 709
P Q Go Back
Full Screen
Close
Quit
Square Construction algorithm
Perpendicular at a
point
Given a line, draw a perpendicular to IIT Delhi
it passing through a given point on it. S. Arun-Kumar, CSE
Assume given a line l containing a Title Page
point X. Contents
JJ II
J I
Page 21 of 709
Go Back
X l
Full Screen
Close
Quit
Solution:
Perpendicular at a IIT Delhi
point S. Arun-Kumar, CSE
Title Page
1. Choose a length c > 0. With X as

Contents
centre mark off points Y and Z on l
JJ II
on either side of X, such that Y X =
c = XZ. Y Z = 2c. J I
Page 22 of 709
2. Draw Circles C1(Y, 2c) and C2(Z, 2c)

Go Back
respectively.
Full Screen
3. Join the points of intersection of the Close
two circles. Quit

Perpendicular at a
Point: 1
Choose a length c > 0. With X as
centre mark off points Y and Z on l IIT Delhi
on either side of X, such that Y X = S. Arun-Kumar, CSE
c = XZ. Y Z = 2c. Title Page
Contents
JJ II
J I
Page 23 of 709
Go Back
Y X Z l
Full Screen
Close
Quit
Perpendicular at a
Point: 2
Draw Circles C1(Y, 2c) and C2(Z, 2c) IIT Delhi
respectively. S. Arun-Kumar, CSE
Title Page
Contents
U
JJ II
J I
Y X Z l Page 24 of 709
Go Back
V Full Screen
Close
Quit
Perpendicular at a
Point: 3
Choose a length c > 0. With X as
centre mark off points Y and Z on l IIT Delhi
on either side of X, such that Y X = S. Arun-Kumar, CSE
c = XZ. Y Z = 2c. Title Page
Contents
JJ II
U
J I
Page 25 of 709
Go Back
Y X Z l
Full Screen
Close
V
Quit
Perpendicular at a
point: Justification IIT Delhi
S. Arun-Kumar, CSE
1. The two circles intersect at points U Title Page
and V on either side of line l. Contents

↔
2. U V is a perpendicular bisector of JJ II
Y Z. J I
3. Since Y X = c = XZ and Y Z = 2c, Page 26 of 709
↔
U V is perpendicular to l and passes Go Back
through X. Full Screen
Close
Back to square 1
Quit
1.2. Our Computing Tool
Previous: Introduction to Computing
IIT Delhi
1. The Digital Computer: Our Computing Tool
2. Algorithms S. Arun-Kumar, CSE
3. Programming Language
Title Page
4. Programs and Languages
5. Programs Contents
6. Programming
JJ II
7. Computing Models
8. Primitives J I
9. Primitive expressions
Page 27 of 709
10. Methods of combination
11. Methods of abstraction
Go Back
12. The Functional Model
13. Mathematical Notation 1: Factorial Full Screen
14. Mathematical Notation 2: Factorial

Close
16. A Functional Program: Factorial Quit
17. A Computation: Factorial
20. A Computation: Factorial IIT Delhi

S. Arun-Kumar, CSE
23. A Computation: Factorial Title Page
24. Standard ML
Contents
25. SML: Important Features
Next: Primitives: Integer & Real JJ II
J I
Page 28 of 709
Go Back
Full Screen
Close
Quit
The Digital Computer:
Our Computing Tool IIT Delhi
Algorithm: A finite specification of the S. Arun-Kumar, CSE
solution to a given problem using the Title Page
primitives of the computing tool. Contents
JJ II
• It specifies a definite input and out-
J I
put
Page 29 of 709
• It is unambiguous Go Back
• It specifies a solution as a finite pro- Full Screen
cess i.e. the number of steps in the Close
computation is finite Quit

Algorithms
An algorithm will be written in a mix- IIT Delhi
ture of English and standard mathe- S. Arun-Kumar, CSE
matical notation. Usually, Title Page
• algorithms written in a natural lan- Contents
guage are often ambiguous JJ II
• mathematical notation is not am- J I
biguous, but still cannot be under- Page 30 of 709
stood by machine Go Back
• algorithms written by us use various Full Screen
mathematical properties. We know Close
them, but the machine doesn’t. Quit

Programming
Language IIT Delhi
• Require a way to communicate with

S. Arun-Kumar, CSE
a machine which has essentially no Title Page
intelligence or understanding. Contents
JJ II
• Translate the algorithm into a form
J I
that may be “understood” by a ma-
Page 31 of 709
chine
Go Back
• This “form” is usually a program Full Screen
Program: An algorithm written in a Close
programming language.
Quit
Programs and
Languages IIT Delhi
• Every programming language has a S. Arun-Kumar, CSE
well defined vocabulary and a well Title Page
defined grammar Contents
• Each program has to be written fol- JJ II
lowing rigid grammatical rules J I
• A programming language and the Page 32 of 709
computer together form our single Go Back
computing tool Full Screen
• Each program uses only the primi- Close
tives of the computing tool Quit

Programs
Program: An algorithm written in the IIT Delhi
grammar of a programming language. S. Arun-Kumar, CSE
Title Page
A grammar is a set of rules for forming Contents
sentences in a language. JJ II
Each programming language also has J I
its own vocabulary and grammar just Page 33 of 709
as in the case of natural languages. Go Back
Full Screen
We will learn the grammar of the lan-

guage as we go along. Close
Quit
Programming
The act of writing programs and test- IIT Delhi
ing them is programming S. Arun-Kumar, CSE
Title Page
Even though most programming lan-

Contents
guages use essentially the same
JJ II
computing primitives, each program-
J I
ming language needs to be learned.
Page 34 of 709
Programming languages differ from Go Back
each other in terms of the conve- Full Screen
nience and facilties they offer even Close
though they are all equally powerful. Quit

Computing Models
We consider mainly two models. IIT Delhi
• Functional: A program is specified S. Arun-Kumar, CSE
simply as a mathematical expres- Title Page
sion Contents
• Imperative: A program is specified JJ II
by a sequence of commands to be J I
executed. Page 35 of 709
Go Back
Programming languages also come

mainly in these two flavours. We Full Screen
will often identify the computing model Close
with the programming language. Quit

Primitives IIT Delhi
Every programming language offers S. Arun-Kumar, CSE
the following capabilities to define and Title Page
use: Contents
• Primitive expressions and data JJ II
• Methods of combination of expres- J I
sions and data Page 36 of 709
• Methods of abstraction of both ex- Go Back
pressions and data Full Screen
The functional model Close
Quit
Primitive expressions
The simplest objects and operations
in the computing model. These in- IIT Delhi
clude S. Arun-Kumar, CSE
• basic data elements: numbers, Title Page
characters, truth values etc. Contents
JJ II
• basic operations on the data ele-
J I
ments: addition, substraction, mul-
Page 37 of 709
tiplication, division, boolean opera-
tions, string operations etc.
Go Back
Full Screen
• a naming mechanism for various Close
quantitities and expressions to be

Quit
used without repeating definitions

IIT Delhi
Methods of S. Arun-Kumar, CSE
combination Title Page
Contents
Means of combining simple expres-
JJ II
sions and objects to obtain more com-
J I
plex expressions and objects.
Page 38 of 709
Examples: composition of functions, Go Back
inductive definitions
Full Screen
Close
Quit
IIT Delhi
Methods of abstraction
S. Arun-Kumar, CSE
Title Page
Means of naming and using groups of Contents
objects and expressions as a single JJ II
unit J I
Examples: functions, data structures, Page 39 of 709
modules, classes etc. Go Back
Full Screen
Close
Quit
The Functional Model
The functional model is very conve- IIT Delhi
nient and easy to use: S. Arun-Kumar, CSE
• Programs are written (more or less) Title Page
in mathematical notation Contents
• It is like using a hand-held calcula- JJ II
tor J I
• Very interactive and so answers are Page 40 of 709
immediately available Go Back
• Very convenient for developing, Full Screen
testing and proving algorithms Close
Quit
Standard ML
IIT Delhi
S. Arun-Kumar, CSE
Mathematical Notation Title Page
1: Factorial Contents
JJ II

1 if n < 1 J I
n! =
n × (n − 1)! otherwise Page 41 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
Mathematical Notation S. Arun-Kumar, CSE
2: Factorial Title Page
Contents
Or more informally, JJ II
J I
1 if n < 1
n! = Page 42 of 709
1 × 2 × . . . × n otherwise
Go Back
Full Screen
Close
Quit
Mathematical Notation
IIT Delhi
3: Factorial
S. Arun-Kumar, CSE
Title Page
How about this? Contents
JJ II
1 if n < 1
n! = J I
(n + 1)!/(n + 1) otherwise Page 43 of 709
Mathematically correct but computa- Go Back
tionally incorrect! Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
A Functional Program: Title Page
Factorial Contents
JJ II
J I
fun fact n = if n < 1 then 1
else n * fact (n-1) Page 44 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
A Computation: Title Page
Factorial JJ
Contents
II
sml
J I
Standard ML of New Jersey,
Page 45 of 709
-
Go Back
Full Screen
Close
Quit
IIT Delhi
A Computation:
S. Arun-Kumar, CSE
Title Page
Factorial Contents
sml JJ II
Standard ML of New Jersey, J I
- fun fact n = Page 46 of 709
= Go Back
Full Screen
Close
Quit
IIT Delhi
A Computation: S. Arun-Kumar, CSE
Factorial Title Page
Contents
sml JJ II
J I
- fun fact n =
Page 47 of 709
= if n < 1 then 1
Go Back
=
Full Screen
Close
Quit
A Computation: IIT Delhi
Factorial S. Arun-Kumar, CSE
Title Page
sml Contents
Standard ML of New Jersey, JJ II

- fun fact n =
J I
= if n < 1 then 1
Page 48 of 709
= else n * fact (n-1);

Go Back
val fact = fn : int -> int
Full Screen
-
Close
Quit
A Computation:
Factorial IIT Delhi
S. Arun-Kumar, CSE
sml Title Page
Standard ML of New Jersey, Contents
- fun fact n = JJ II
= if n < 1 then 1
J I
Page 49 of 709

Go Back
- fact 8;
Full Screen
val it = 40320 : int
Close
-
Quit
A Computation:
Factorial IIT Delhi
sml S. Arun-Kumar, CSE
Standard ML of New Jersey, Title Page
- fun fact n = Contents
= if n < 1 then 1 JJ II
J I
Page 50 of 709
- fact 8;
Go Back
val it = 40320 : int
Full Screen
- fact 9;
Close
val it = 362880 : int
- Quit
A Computation:
Factorial
sml IIT Delhi
Standard ML of New Jersey, S. Arun-Kumar, CSE
- fun fact n = Title Page
= if n < 1 then 1 Contents
= else n * fact (n-1); JJ II
val fact = fn : int -> int J I
- fact 8; Page 51 of 709
val it = 40320 : int Go Back
- fact 9; Full Screen
val it = 362880 : int Close
- Quit
More SML
IIT Delhi
Standard ML S. Arun-Kumar, CSE
• Originated as part of a theorem- Title Page
proving development project Contents
• Runs on both Windows and UNIX JJ II
environments J I
Page 52 of 709
• Is free!
Go Back
• http://www.smlnj.org Full Screen
Close
Quit
SML: Important
Features IIT Delhi
• Has a small vocabulary of just a few S. Arun-Kumar, CSE
short words Title Page
• Far more “intelligent” than currently Contents
available languages: JJ II
J I
– automatically finds out what vari-
Page 53 of 709
ous names mean and

Go Back
– their correct usage

Full Screen
• Haskell, Miranda and Caml are a Close
few other such languages. Quit

1.3. Primitives: Integer & Real
Previous: Primitives: Integer & Real
IIT Delhi
1. Algorithms & Programs
2. SML: Primitive Integer Operations 1 S. Arun-Kumar, CSE
3. SML: Primitive Integer Operations 1

Title Page
5. SML: Primitive Integer Operations 1 Contents

JJ II
8. SML: Primitive Integer Operations 2 J I
Page 54 of 709
Go Back
13. SML: Primitive Integer Operations 3 Full Screen

Close
16. SML: Primitive Integer Operations 3 Quit
18. Quotient & Remainder
19. SML: Primitive Real Operations 1
20. SML: Primitive Real Operations 1 IIT Delhi

S. Arun-Kumar, CSE
23. SML: Primitive Real Operations 3 Title Page

Contents
25. SML: Precision
26. Fibonacci Numbers JJ II
27. Euclidean Algorithm
J I
Next: Technical Completeness & Algorithms
Page 55 of 709
Go Back
Full Screen
Close
Quit
Algorithms & Programs IIT Delhi
• Algorithm S. Arun-Kumar, CSE
• Need for a formal notation Title Page
• Programs Contents
JJ II
• Programming languages
J I
• Programming Page 56 of 709
• Functional Programming Go Back
• Standard ML Full Screen
Factorial Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
SML: Primitive Integer Title Page
Operations 1 Contents
JJ II
sml J I
Standard ML of New Jersey, Page 57 of 709
- Go Back
Full Screen
Close
Quit
IIT Delhi
SML: Primitive Integer S. Arun-Kumar, CSE
Operations 1 Title Page
Contents
sml JJ II
Standard ML of New Jersey, J I
- val x = 5; Page 58 of 709
val x = 5 : int Go Back
-
Full Screen
Close
Quit
SML: Primitive Integer IIT Delhi
Operations 1 S. Arun-Kumar, CSE
Title Page
sml Contents
Standard ML of New Jersey, JJ II
- val x = 5; J I
val x = 5 : int Page 59 of 709
- val y = 6; Go Back
val y = 6 : int
Full Screen
-
Close
Quit
SML: Primitive Integer
Operations 1 IIT Delhi
S. Arun-Kumar, CSE
sml Title Page
Standard ML of New Jersey, Contents
- val x = 5; JJ II
val x = 5 : int J I
- val y = 6; Page 60 of 709
val y = 6 : int Go Back
- x+y;
Full Screen
val it = 11 : int
Close
-
Quit
Operations 1 IIT Delhi
- val x = 5; Contents
val x = 5 : int JJ II
- val y = 6; J I
val y = 6 : int Page 61 of 709
- x+y; Go Back
val it = 11 : int
Full Screen
- x-y;
Close
val it = ˜1 : int
Quit
-
Operations 1
IIT Delhi
S. Arun-Kumar, CSE
- val x = 5;
val x = 5 : int Title Page
- val y = 6; Contents
val y = 6 : int JJ II
- x+y; J I
val it = 11 : int Page 62 of 709
- x-y; Go Back
val it = ˜1 : int Full Screen
- it + 5; Close
val it = 4 : int Quit
-
Operations 2
IIT Delhi
val x = 5 : int
S. Arun-Kumar, CSE
- val y = 6;
val y = 6 : int Title Page
- x+y; Contents
val it = 11 : int JJ II
- x-y; J I
val it = ˜1 : int Page 63 of 709
- it + 5; Go Back
val it = 4 : int Full Screen
- x * y; Close
-
Operations 2
IIT Delhi
val y = 6 : int
S. Arun-Kumar, CSE
- x+y;
val it = 11 : int Title Page
- x-y; Contents
val it = ˜1 : int JJ II
- it + 5; J I
- x * y; Go Back
- val a = 25; Close
val a = 25 : int Quit
-
Operations 2
IIT Delhi
val it = 11 : int
S. Arun-Kumar, CSE
- x-y;
val it = ˜1 : int Title Page
- it + 5; Contents
- x * y; J I
- val a = 25; Go Back
val a = 25 : int Full Screen
- val b = 7; Close
val b = 7 : int Quit
-
Operations 2
IIT Delhi
val it = ˜1 : int
S. Arun-Kumar, CSE
- it + 5;
val it = 4 : int Title Page
- x * y; Contents
- val a = 25; J I
val a = 25 : int Page 66 of 709
- val b = 7; Go Back
val b = 7 : int Full Screen
- val q = a div b; Close
val q = 3 : int Quit
-
Operations 2
- x * y; IIT Delhi
val it = 30 : int S. Arun-Kumar, CSE
- val a = 25; Title Page
val a = 25 : int Contents
- val b = 7; JJ II
val b = 7 : int J I
- val q = a div b; Page 67 of 709
val q = 3 : int Go Back
- val r = a mod b; Full Screen
GC #0.0.0.0.2.45: (0 ms) Close
val r = 4 : int Quit
-
Operations 3
IIT Delhi
- val a = 25;
S. Arun-Kumar, CSE
val a = 25 : int
- val b = 7; Title Page
val b = 7 : int Contents
- val q = a div b; JJ II
val q = 3 : int J I
- val r = a mod b; Page 68 of 709
GC #0.0.0.0.2.45: (0 ms) Go Back
val r = 4 : int Full Screen
- a = b*q + r; Close
val it = true : bool Quit
-
Operations 3
IIT Delhi
- val b = 7;
S. Arun-Kumar, CSE
val b = 7 : int
- val q = a div b; Title Page
val q = 3 : int Contents
- val r = a mod b; JJ II
GC #0.0.0.0.2.45: (0 ms) J I
val r = 4 : int Page 69 of 709
- a = b*q + r; Go Back
val it = true : bool Full Screen
- val c = ˜7; Close
val c = ˜7 : int Quit
-
Operations 3
IIT Delhi
- val q = a div b;
S. Arun-Kumar, CSE
val q = 3 : int
- val r = a mod b; Title Page
GC #0.0.0.0.2.45: (0 ms) Contents
val r = 4 : int JJ II
- a = b*q + r; J I
val it = true : bool Page 70 of 709
- val c = ˜7; Go Back
val c = ˜7 : int Full Screen
- val q1 = a div c; Close
val q1 = ˜4 : int Quit
-
Operations 3
IIT Delhi
- val r = a mod b;
S. Arun-Kumar, CSE
GC #0.0.0.0.2.45: (0 ms)
val r = 4 : int Title Page
- a = b*q + r; Contents
val it = true : bool JJ II
- val c = ˜7; J I
val c = ˜7 : int Page 71 of 709
- val q1 = a div c; Go Back
val q1 = ˜4 : int Full Screen
- val r1 = a mod c; Close
val r1 = ˜3 : int Quit
-
Operations 3
IIT Delhi
val r = 4 : int
S. Arun-Kumar, CSE
- a = b*q + r;
val it = true : bool Title Page
- val c = ˜7; Contents
val c = ˜7 : int JJ II
- val q1 = a div c; J I
val q1 = ˜4 : int Page 72 of 709
- val r1 = a mod c; Go Back
val r1 = ˜3 : int Full Screen
- a = c*q1 + r1; Close
val it = true : bool Quit
-
Quotient & Remainder
IIT Delhi
For any two integers a and b, the quo-

S. Arun-Kumar, CSE
tient q and remainder r are uniquely
determined to satisfy Title Page
• Contents
JJ II
a=b×q+r
J I
• Page 73 of 709

0 ≤ r 0 Go Back
b < r ≤ 0 when b < 0 Full Screen
So 0 ≤ |r| < |b| always. Close
Quit
IIT Delhi
SML: Primitive Real S. Arun-Kumar, CSE
Operations 1 Title Page
Contents
sml JJ II
J I
- val real_a = real a;
Page 74 of 709
val real_a = 25.0 : real

Go Back
-
Full Screen
Close
Quit
SML: Primitive Real
Operations 1
IIT Delhi
- val real_a = real a; Contents
val real_a = 25.0 : real JJ II

- real_a + b;
J I
stdIn:40.1-40.11 Error: operator and
Page 75 of 709
operator domain: real * real

Go Back
operand: real * int

Full Screen
in expression:
Close
real_a + b
- Quit
SML: Primitive Real
Operations 1
stdIn:40.1-40.11 Error: operator and IIT Delhi
operator domain: real * real S. Arun-Kumar, CSE
operand: real * int Title Page
in expression: Contents
real_a + b JJ II
- b + real_a; J I
stdIn:1.1-2.6 Error: operator and ope

Page 76 of 709
operator domain: int * int Go Back
operand: int * real Full Screen
in expression: Close
b + real_a Quit
-
SML: Primitive Real
Operations 2
- val a = 25.0; IIT Delhi
val a = 25.0 : real S. Arun-Kumar, CSE
- val b = 7.0; Title Page
val b = 7.0 : real Contents
- a/b; JJ II
val it = 3.57142857143 : real J I
- a div b; Page 77 of 709
stdIn:49.3-49.6 Error: overloaded var Go Back
symbol: div Full Screen
type: real Close
GC #0.0.0.0.3.98: (0 ms) Quit
-
SML: Primitive Real IIT Delhi
Operations 3 S. Arun-Kumar, CSE
- val c = a/b; Title Page
val c = 3.57142857143 : real Contents
- trunc(c); JJ II
val it = 3 : int J I
- trunc (c + 0.5); Page 78 of 709
val it = 4 : int Go Back
- Full Screen
Close
Quit
SML: Primitive Real
Operations 4
- val d = 3.0E10; IIT Delhi
val d = 30000000000.0 : real S. Arun-Kumar, CSE
- val pi = 0.314159265E1; Title Page
val pi = 3.14159265 : real Contents
- - d+pi; JJ II
val it = 30000000003.1 : real J I
- d-pi; Page 79 of 709
val it = 29999999996.9 : real Go Back
- pi + d; Full Screen
val it = 30000000003.1 : real Close
- Quit
SML: Precision
- pi + d*10.0; IIT Delhi
val it = 300000000003.0 : real S. Arun-Kumar, CSE
- pi + d*100.0;
Title Page
val it = 3E12 : real

Contents
- d*100.0 + pi;
JJ II
J I
- d*100.0 -pi;
Page 80 of 709
- d*10.0 - pi; Go Back
val it = 299999999997.0 : real Full Screen
- Close
Quit
1.4. Example: Fibonacci
1. Fibonacci Numbers: 1
3. Fibonacci Numbers: 3 IIT Delhi
S. Arun-Kumar, CSE
6. Is Fa(n, 1, 1) = F(n)? Title Page
7. Trial & Error

Contents
8. Generalization
9. Proof JJ II
10. Another Generalization
11. Try Proving it! J I
12. Another Generalization

Page 81 of 709
13. Try Proving it!
14. Complexity Go Back
15. Complexity
Full Screen
16. Time Complexity: R
17. Time Complexity Close
18. Time Complexity: R

Quit
19. Bound on R
20. Other Bounds: CF
21. Other Bounds: AF

IIT Delhi
S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 82 of 709
Go Back
Full Screen
Close
Quit
Fibonacci Numbers: 1 IIT Delhi

 F(0) = 1 S. Arun-Kumar, CSE
F(1) = 1 Title Page
F(n) = F(n − 1) + F(n − 2) if n > 1


Contents
JJ II
fun fib (n) = J I
if (n = 0) orelse (n = 1) Page 83 of 709
then 1 Go Back
else fib (n-1) + fib (n-2); Full Screen
Close
Quit
Fibonacci Numbers: 2 IIT Delhi
 S. Arun-Kumar, CSE
 F(0) = 1
F(1) = 1 Title Page
F(n) = F(n − 1) + F(n − 2) if n > 1 Contents


JJ II
Alternatively, J I

1 if n = 0 Page 84 of 709
F(n) = 1 if n = 1 Go Back
F(n − 1) + F(n − 2) if n > 1


Full Screen
Close
Quit
Fibonacci Numbers: 3
fun fib (n) = IIT Delhi
if (n = 0) orelse (n = 1) S. Arun-Kumar, CSE
then 1 Title Page
else fib (n-1) + fib (n-2); Contents
JJ II
Alternatively, J I
fun fib (n) = Page 85 of 709
if (n = 0) then 1 Go Back
else if (n = 1) then 1 Full Screen
else fib (n-1) + fib (n-2); Close
Quit
Fibonacci Numbers: 4

 F(0) = 1 IIT Delhi
F(1) = 1 S. Arun-Kumar, CSE
F(n) = F(n − 1) + F(n − 2) if n > 1


Title Page
Alternatively, Contents
JJ II
F(n) = Fa(n, 1, 1) J I
where Page 86 of 709
 Go Back
a if n = 0
Fa(n, a, b) = b if n = 1 Full Screen
Fa(n − 1, b, a + b) if n > 1 Close


Quit
IIT Delhi
Fibonacci Numbers: 5 S. Arun-Kumar, CSE
Title Page
fun fib_a (n, a, b) =

Contents
if (n = 0) then a
JJ II
else if (n = 1) then b
J I
else fib_a (n, b, a+b);
Page 87 of 709
fun fib (n) = fib_a (n, 1, 1); Go Back
Full Screen
Close
Quit
Is Fa(n, 1, 1) = F(n)? IIT Delhi
S. Arun-Kumar, CSE
Intuition Fa is a generalization of F. Title Page
Question 1 What does it actually gen- Contents
eralize to? JJ II
Question 2 Does the generalization J I
have a non-recursive form? Page 88 of 709
Trial & Error Can we use trial and er- Go Back
ror to get a non-recursive form? Full Screen
Close
Quit
Trial & Error IIT Delhi
S. Arun-Kumar, CSE
Fa(2, a, b) = a + b
Title Page
Fa(3, a, b) = Fa(2, b, a + b)
Contents
= a + 2b
JJ II
Fa(4, a, b) = Fa(3, b, a + b)
J I
= Fa(2, a + b, a + 2b)
= 2a + 3b Page 89 of 709
Fa(5, a, b) = Fa(2, a + 2b, 2a + 3b) Go Back
= 3a + 5b Full Screen
Close
Quit
Generalization
• Fa(0, a, b) = a IIT Delhi
S. Arun-Kumar, CSE
• Fa(1, a, b) = b
Title Page
• Fa(n, a, b) = aF(n − 2) + bF(n − 1) Contents
• When a = 1 and b = 1, for all n ≥ 0, JJ II
Fa(n, a, b) = F(n) J I
Theorem 1 For all integers a, b and Page 90 of 709
n > 1, Go Back
Full Screen
Fa(n, a, b) = aF(n − 2) + bF(n − 1) Close
Quit
Proof by Induction on
n>1
Proof:
Basis For n = 2, Fa(2, a, b) = a + b = IIT Delhi
aF(0) + bF(1) S. Arun-Kumar, CSE
Title Page
Induction hypothesis (IH) Assume
Fa(k, a, b) = aF(k − 2) + bF(k − 1), Contents
for some k > 1 and all integers a, b JJ II
Induction Step J I
Page 91 of 709
Fa(k + 1, a, b)
Go Back
= Fa(k, b, a + b) Definition of Fa
= bF(k − 2) + (a + b)F(k − 1) IH Full Screen
= aF(k − 1) + b(F(k − 2) + F(k − 1)) Close
= aF(k − 1) + bF(k) Definition of F Quit

IIT Delhi
S. Arun-Kumar, CSE
Another Generalization Title Page
Try to prove a different and more di- Contents
rect theorem. JJ II
J I
Theorem 2 For all integers n > 1,
Fa(n, 1, 1) = F(n) Page 92 of 709
Go Back
Full Screen
Close
Quit
Try Proving it!
Proof: By induction on n > 1.
IIT Delhi
Basis For n = 0 and n = 1, Fa(0, 1, 1) =

S. Arun-Kumar, CSE
1 = Fa(1)
Title Page
Induction hypothesis (IH) Assume Contents
Fa(k, 1, 1) = F(k), for some k > 1 JJ II

Induction Step
J I
Fa(k + 1, 1, 1) Page 93 of 709
= Fa(k, 1, 2) Definition of Fa Go Back
= ? ? ? IH Full Screen
STUCK! Close
Quit

Another Generalization
Try to prove a different and more di- IIT Delhi
rect theorem. S. Arun-Kumar, CSE
Theorem 3 For all integers n ≥ 1 and Title Page
j ≥ 1, Contents
JJ II
Fa(n, F(j − 1), F(j)) = F(n + j − 1) J I
Page 94 of 709
Proof: By induction on n ≥ 1, for all

Go Back
values of j ≥ 1.
Full Screen
Corollary 4 For all integers n ≥ 1, Close
Fa(n, F(0), F(1)) = F(n)

Quit
Try Proving it!
Basis For n = 1, Fa(1, F(j − 1), F(j)) =
F(j) IIT Delhi
Induction hypothesis (IH) For some S. Arun-Kumar, CSE
k > 1 and all j ≥ 1, Title Page
Fa(k, F(j − 1), F(j)) = F(k + j − 1) Contents
Induction Step We need to prove JJ II
Fa(k + 1, F(j − 1), F(j)) = F(k + j). J I
Page 95 of 709
Fa(k + 1, F(j − 1), F(j)) Go Back
= Fa(k, F(j), F(j − 1) + F(j))

Full Screen
= Fa(k, F(j), F(j + 1))

Close
= Fa(k + (j + 1) − 1) IH
Quit
= Fa(k + j)
IIT Delhi
Complexity S. Arun-Kumar, CSE
Title Page
• Time complexity:
Contents
– No of additions: AF(n) JJ II
– No of comparisons: CF(n) J I
– No of recursive calls to F: RF(n) Page 96 of 709
• Space complexity: Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Complexity Title Page
• Time complexity: Contents
• Space complexity: JJ II
J I
– left-to-right evaluation: LRF(n)
Page 97 of 709
– arbitrary evaluation: UF(n) Go Back
Full Screen
Close
Quit
Time Complexity:
IIT Delhi
R S. Arun-Kumar, CSE
• Hardware operations like addition Title Page
and comparisons are usually very Contents
fast compared to software opera- JJ II
tions like recursion unfolding J I
• The number of recursion unfoldings Page 98 of 709
also includes comparisons and ad- Go Back
ditions. Full Screen
Close
Quit
Time Complexity IIT Delhi
S. Arun-Kumar, CSE
• It is enough to put bounds on the

Title Page
number of recursion unfoldings and

Contents
not worry about individual hardware
JJ II
operations.
J I
• Similar theorems may be proved for
Page 99 of 709
any operation by counting and in-

Go Back
duction.
So we concentrate on R.
Full Screen
Close
Quit
Time Complexity: R IIT Delhi
• RF(0) = RF(1) = 0 S. Arun-Kumar, CSE
• RF(n) = 2 + RF(n − 1) + RF(n − 2) Title Page
for n > 1 Contents
JJ II
To solve the equation as initial value
J I
problem and obtain an upper bound
Page 100 of 709
we guess the following theorem.
Go Back
Theorem 5 RF(n) ≤ 2n−1 for all n > 2 Full Screen
Proof: By induction on n > 2. Close
Quit
Bound on R
Basis n = 3. RF(3) = 2 + 2 + 0 ≤ 23−1 IIT Delhi
Induction hypothesis (IH) For some S. Arun-Kumar, CSE
k > 2, RF(k) ≤ 2k−1 Title Page
Induction Step If n = k + 1 then n > 3 Contents
JJ II
RF(n)
J I
= 2 + RF(n − 2) + RF(n − 1)
2 + 2n−3 + 2n−2 (IH)
Page 101 of 709
≤
2.2n−3 + 2n−2 for n > 3, 2n−3 ≥ 2
Go Back
≤
2n−2 + 2n−2
Full Screen
=
2n−1
Close
=
Quit
IIT Delhi
Other Bounds: CF S. Arun-Kumar, CSE
Title Page
One comparison for each call.

Contents
• CF(0) = CF(1) = 1 JJ II
• CF(n) = 1 + CF(n − 1) + CF(n − 2) for J I
n>1 Page 102 of 709
Theorem 6 CF(n) ≤ 2n for all n ≥ 0. Go Back
Full Screen
Close
Quit
Other Bounds: AF IIT Delhi
S. Arun-Kumar, CSE
No additions for the basis and one ad- Title Page
dition in each call. Contents
• AF(0) = AF(1) = 0 JJ II
• AF(n) = 1 + AF(n − 1) + AF(n − 2) J I
for n > 1 Page 103 of 709
Theorem 7 AF(n) ≤ 2n−1 for all n > Go Back
0. Full Screen
Close
Quit
1.5. Primitives: Booleans
1. Boolean Conditions
IIT Delhi
2. Booleans in SML
3. Booleans in SML S. Arun-Kumar, CSE
4. ∧ vs. andalso
Title Page
5. ∨ vs. orelse
6. SML: orelse Contents
7. SML: andalso
JJ II
8. and, andalso, ⊥
9. or, orelse, ⊥ J I
10. Complex Boolean Conditions
Page 104 of 709
Go Back
Full Screen
Close
Quit
Boolean Conditions IIT Delhi
• Two (truth) value set : {true, S. Arun-Kumar, CSE
false} Title Page
• Boolean conditions are those state- Contents
ments or names which can take JJ II
only truth values. J I
Examples: n < 0, true, Page 105 of 709
• Negation operator: not Go Back
Examples: not (n < 0), not Full Screen
true, not false Close
Quit
Booleans in SML
- val tt = true; IIT Delhi
val tt = true : bool S. Arun-Kumar, CSE
- not(tt); Title Page
val it = false : bool Contents
- val n = 10; JJ II
val n = 10 : int J I
- n < 10; Page 106 of 709
val it = false : bool Go Back
- not (n<10); Full Screen
val it = true : bool

Close
-
Quit
Booleans in SML IIT Delhi
Examples: S. Arun-Kumar, CSE
- (n >= 10) andalso (n=10); Title Page
val it = true : bool Contents
- n < 0 orelse n >= 10; JJ II

- not ((n >= 10) J I
= andalso (n=10)) Page 107 of 709
= orelse n < 0 orelse n >= 10; Go Back
- Close
Quit
IIT Delhi
vs. andalso
S. Arun-Kumar, CSE
∧ Title Page
p q p∧q p andalso q Contents
true true true true JJ II
true false false false J I

false true false false Page 108 of 709
false false false false

Go Back
Full Screen
Close
Quit
IIT Delhi
vs. orelse
S. Arun-Kumar, CSE
∨ Title Page
p q p∨q p orelse q Contents
true true true true JJ II
true false true true J I

false true true true Page 109 of 709
false false false false

Go Back
Full Screen
Close
Quit
SML: orelse
Standard ML of New Jersey, IIT Delhi
- val tt = true; S. Arun-Kumar, CSE
val tt = true : bool Title Page
- val ff = false;
val ff = false : bool Contents
- fun gtz n = if n=1 then true JJ II
= else gtz (n-1); J I
val gtz = fn : int -> bool Page 110 of 709
- tt orelse (gtz 0); Go Back
- Close
Quit
SML: andalso IIT Delhi
- (gtz 0) orelse tt; S. Arun-Kumar, CSE
Title Page
Interrupt Contents
- ff andalso (gtz 0);

val it = false : bool JJ II
- (gtz 0) andalso ff; J I
Page 111 of 709
Interrupt Go Back
- Full Screen
Close
Quit
IIT Delhi
and, andalso, ⊥ S. Arun-Kumar, CSE
p q p∧q p andalso q Title Page
true ⊥ ⊥ ⊥ Contents
⊥ true ⊥ ⊥ JJ II
false ⊥ false false J I
⊥ false false ⊥ Page 112 of 709
∧ is commutative whereas andalso Go Back
is not. Full Screen
Close
Quit
IIT Delhi
or, orelse, ⊥ S. Arun-Kumar, CSE
p q p∨q p orelse q Title Page
true ⊥ true true Contents
⊥ true true ⊥ JJ II
false ⊥ ⊥ ⊥ J I
⊥ false ⊥ ⊥ Page 113 of 709
∨ is commutative whereas orelse is Go Back
not. Full Screen
Close
Quit
Complex Boolean IIT Delhi
Conditions S. Arun-Kumar, CSE
Title Page
Assume p and q are boolean condi- Contents
tions JJ II
J I
p orelse q ≡ if p then true else q Page 114 of 709
Go Back
p andalso q ≡ if p then q else false Full Screen
Close
Quit
2. Algorithms: Design & Refinement
2.1. Technical Completeness & Algorithms IIT Delhi
1. Recapitulation: Integers & Real

S. Arun-Kumar, CSE
2. Recap: Integer Operations
3. Recapitulation: Real Operations Title Page
4. Recapitulation: Simple Algorithms

Contents
5. More Algorithms
6. Powering: Math JJ II
7. Powering: SML
J I
8. Technical completeness
9. What SML says Page 115 of 709

Go Back
11. What SML says ... contd
12. Powering: Math 1

Full Screen
13. Powering: SML 1
14. Technical Completness Close
15. What SML says

Quit
16. Powering: Integer Version
17. Exceptions: A new primitive
18. Integer Power: SML
19. Integer Square Root 1 IIT Delhi
20. Integer Square Root 2

S. Arun-Kumar, CSE
21. An analysis
22. Algorithmic idea Title Page
23. Algorithm: isqrt

Contents
24. Algorithm: shrink
25. SML: shrink JJ II
26. SML: intsqrt
27. Run it! J I
28. SML: Reorganizing Code

Page 116 of 709
29. Intsqrt: Reorganized
30. shrink: Another algorithm Go Back
31. Shrink2: SML

Full Screen
32. Shrink2: SML ... contd
Close
Quit
IIT Delhi
Recapitulation: S. Arun-Kumar, CSE
Integers & Real Title Page
Contents
• Primitive Integer Operations JJ II
• Primitive Real Operations J I
• Some algorithms Page 117 of 709
Go Back
Forward
Full Screen
Close
Quit
Recap: Integer IIT Delhi
Operations S. Arun-Kumar, CSE
• Primitive Integer Operations Title Page
– Naming, +, −, ∼
Contents
JJ II
– Multiplication, division
J I
– Quotient & remainder
Page 118 of 709
• Primitive Real Operations Go Back
• Some algorithms Full Screen
Back Close
Quit
Recapitulation: Real
Operations IIT Delhi
• Primitive Integer Operations S. Arun-Kumar, CSE
• Primitive Real Operations Title Page
Contents
– Integer to Real
JJ II
– Real to Integer J I
– Real addition & subtraction Page 119 of 709
– Real division Go Back
– Real Precision Full Screen
• Some algorithms Close
Back Quit
Recapitulation: Simple IIT Delhi
Algorithms S. Arun-Kumar, CSE
• Primitive Integer Operations Title Page
• Primitive Real Operations

Contents
JJ II
• Some algorithms J I
– Factorial Page 120 of 709
– Fibonacci Go Back
– Euclidean GCD Full Screen
Back Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
More Algorithms Title Page
Contents
• Powering JJ II
• Integer square root J I
• Combinations nCk Page 121 of 709
Go Back
Full Screen
Close
Quit
Powering: Math IIT Delhi
For any integer or real number x 6= 0 S. Arun-Kumar, CSE
and non-negative integer n Title Page
xn = |x × x ×{z· · · × x} Contents
n times JJ II
Noting that x0 = 1 we give an induc- J I
tive definition: Page 122 of 709
Go Back

n 1 if n = 0
x =
xn−1 × x otherwise
Full Screen
Close
Quit
IIT Delhi
Powering: SML S. Arun-Kumar, CSE
Title Page
fun power (x:real, n) =

Contents
if n = 0
JJ II
then 1.0
J I
else power (x, n-1) * x
Page 123 of 709
Is it technically complete? Go Back
Full Screen
Close
Quit
Technical IIT Delhi
completeness S. Arun-Kumar, CSE
Title Page
Can it be always guaranteed that
Contents
• x will be real? JJ II
• n will be integer? J I
• n will be non-negative? Page 124 of 709
• x 6= 0? Go Back
If x = 0 what is 0.00?
Full Screen
Close
Quit
What SML says
sml
IIT Delhi
Standard ML of New Jersey
- use "/tmp/power.sml"; S. Arun-Kumar, CSE
[opening /tmp/power.sml] Title Page
val power = fn : real * int -> Contents
real JJ II
val it = () : unit J I
Page 125 of 709
Go Back
Can it be always guaranteed that Full Screen
• x will be real? YES Close
Quit
• n will be integer? YES

Technical IIT Delhi
completeness S. Arun-Kumar, CSE
Can it be always guaranteed that Title Page
• n will be non-negative? NO
Contents
JJ II
• x 6= 0? NO J I
If x = 0 what is 0.00? Page 126 of 709
Go Back
- power(0.0, 0); Full Screen
val it = 1.0 : real Close
Quit
What SML says ...
contd IIT Delhi
Standard ML of New Jersey Title Page
val power = fn : real * int -> real Contents
val it = () : unit
JJ II
- power(˜2.5, 0);
J I
val it = 1.0 : real
Page 127 of 709
- power (0.0, 3);

Go Back
val it = 0.0 : real
Full Screen
- power (2.5, ˜3)
Close
Goes on forever! Quit

IIT Delhi
Powering: Math 1
S. Arun-Kumar, CSE
Title Page
For any real number x and integer n Contents
 1.0/x−n if n < 0
 JJ II
xn = 1 if n = 0 J I
 n−1
x × x otherwise Page 128 of 709
Go Back
Full Screen
Close
Quit
Powering: SML 1 IIT Delhi
fun power (x, n) = S. Arun-Kumar, CSE
if n < 0 Title Page
then 1.0/power(x, ˜n) Contents
else if n = 0 JJ II
then 1.0 J I
else power (x, n-1) * x Page 129 of 709
Go Back
Is this definition technically com- Full Screen
plete? Close
Quit
Technical Completness IIT Delhi
• 0.00 = 1.0 whereas 0.0n = 0 for pos- S. Arun-Kumar, CSE
itive n Title Page
Contents
• What if x = 0.0 and n = −m < 0?

JJ II
Then
0.0n J I
= 1.0/(0.0m) Page 130 of 709
= 1.0/0.0 Go Back
Full Screen
Division by zero!
Close
Quit
What SML says
- power (2.5, ˜2); IIT Delhi
val it = 0.16 : real S. Arun-Kumar, CSE
- power (˜2.5, ˜2); Title Page
val it = 0.16 : real Contents
- power (0.0, 2); JJ II
val it = 0.0 : real J I

- power (0.0, ˜2); Page 131 of 709
val it = inf : real

Go Back
-
Full Screen
Close
SML is somewhat more understand-
ing than most languages Quit
Powering: Integer
Version IIT Delhi

 undefined if n < 0 Title Page
undefined if x = 0&n = 0

n
x = Contents
 1 if x 6= 0&n = 0 JJ II
 n−1

x ×x otherwise J I
Technical completeness requires us Page 132 of 709
to consider the case n < 0. Other- Go Back
wise, the computation can go on for- Full Screen
ever Close
Notation: ⊥ denotes the undef ined

Quit
Exceptions: A new
primitive IIT Delhi
exception negExponent; S. Arun-Kumar, CSE
exception zeroPowerZero; Title Page
fun intpower (x, n) = Contents
if n < 0 JJ II
then raise negExponent J I
else if n = 0 Page 133 of 709
then if x=0 Go Back
then raise zeroPowerZero

Full Screen
else 1
Close
else intpower (x, n-1) * x

Quit
Integer Power: SML
- intpower(3, 4);
val it = 81 : int IIT Delhi
- intpower(˜3, 5); S. Arun-Kumar, CSE
val it = ˜243 : int Title Page
- intpower(3, ˜4); Contents
JJ II
uncaught exception negExponent J I
raised at: intpower.sml:4.16-4.32 Page 134 of 709
- intpower (0, 0); Go Back
Full Screen
uncaught exception zeroPowerZero Close
raised at: stdIn:24.26-24.39 Quit
Back to More Algos

Integer Square Root 1
√
isqrt(n) = b nc IIT Delhi
S. Arun-Kumar, CSE
- fun isqrt n = Title Page
trunc (Real.Math.sqrt Contents
(real (n))); JJ II
val isqrt = fn : int -> int J I

- isqrt (38); Page 135 of 709
val it = 6 : int
Go Back
- isqrt (˜38);
Full Screen
uncaught exception domain error

Close
raised at: boot/real64.sml:89.32-89
Quit
-
Integer Square Root 2
Suppose Real.Math.sqrt were not IIT Delhi
available to us! S. Arun-Kumar, CSE
isqrt(n) of a non-negative integer n is Title Page
the integer k ≥ 0 such that k 2 ≤ n < Contents
(k + 1)2 JJ II
That is, J I

⊥ if n < 0 Page 136 of 709
isqrt(n) =
k otherwise Go Back
Full Screen
where 0 ≤ k 2 ≤ n < (k + 1)2. Close
This value of k is unique!

Quit
An analysis IIT Delhi
S. Arun-Kumar, CSE
0 ≤ k 2 ≤√n < (k + 1)2 Title Page
⇒ 0≤k ≤ n<k+1 Contents
⇒ 0≤k≤n JJ II
Strategy. Use this fact to close in on J I
the value of k. Start with the inter- Page 137 of 709
val [l, u] = [0, n] and try to shrink it till Go Back
it collapses to the interval [k, k] which

contains a single value. Full Screen
Close
Quit
Algorithmic idea
If n = 0 then isqrt(n) = 0. IIT Delhi
Otherwise with [l, u] = [0, n] and S. Arun-Kumar, CSE
l2 ≤ n < u2
Title Page
Contents
use one or both of the following to JJ II
shrink the interval [l, u]. J I
• if (l + 1)2 ≤ n then try [l + 1, u] Page 138 of 709
otherwise l2 ≤ n < (l + 1)2 and k = l Go Back
• if u2 > n then try [l, u − 1] Full Screen
otherwise (u − 1)2 ≤ n < u2 and Close
k =u−1 Quit
IIT Delhi
S. Arun-Kumar, CSE
Algorithm: isqrt Title Page
 Contents
⊥ if n < 0
JJ II
isqrt(n) = 0 if n = 0
J I
shrink(n, 0, n) if n > 0

Page 139 of 709
where Go Back
Full Screen
Close
Quit
Algorithm: shrink
shrink(n, l, u) = IIT Delhi

 l if l = u
 Title Page


 shrink(n, l + 1, u) if l n

l


J I
shrink(n, l, u − 1) if l n
 Page 140 of 709


 Go Back
u−1 if l u
Quit
SML: shrink
exception intervalError; IIT Delhi
fun shrink (n, l, u) = S. Arun-Kumar, CSE
if l>u orelse Title Page
l*l > n orelse Contents
u*u < n JJ II
then raise intervalError J I
else if (l+1)*(l+1) <= n Page 141 of 709
then shrink (n, l+1, u) Go Back
else l; Full Screen
Close
intsqrt
Quit
SML: intsqrt IIT Delhi
S. Arun-Kumar, CSE
exception negError;
Title Page
fun intsqrt n =
Contents
if n<0
JJ II
then raise negError
J I
else if n=0
then 0 Page 142 of 709
else shrink (n, 0, n) Go Back
Full Screen
shrink Close
Quit
Run it!
exception intervalError
IIT Delhi
val shrink =
S. Arun-Kumar, CSE
fn : int * int * int -> int
Title Page
exception negError
val intsqrt = fn : int -> int Contents
val it = () : unit JJ II
- intsqrt 8; J I
- intsqrt 16; Go Back
- intsqrt 99; Close

SML: Reorganizing IIT Delhi
Code S. Arun-Kumar, CSE
Title Page
• shrink was used to develop Contents
intsqrt JJ II
• Is shrink general-purpose J I
enough to be kept separate? Page 144 of 709
• Shouldn’t shrink be placed within Go Back
intsqrt? Full Screen
Close
Quit
Intsqrt: Reorganized
IIT Delhi
exception negError; S. Arun-Kumar, CSE
fun intsqrt n = Title Page
let fun shrink (n, l, u) = ... Contents
in if n<0 JJ II
then raise negError
J I
else if n=0
Page 145 of 709
then 0
Go Back
else shrink (n, 0, n)
Full Screen
end
Close
Quit
shrink: Another
algorithm IIT Delhi
Shrink
S. Arun-Kumar, CSE
shrink2(n, l, u) = Title Page
 Contents

 l if l = u or u = l + 1 JJ II


 shrink2(n, m, u) if l n
 Full Screen
⊥ if l > u

Close
where m = (l + u) div 2 Quit

Shrink2: SML IIT Delhi
S. Arun-Kumar, CSE
fun shrink2 (n, l, u) =

Title Page
if l>u orelse
Contents
l*l > n orelse

JJ II
u*u < n
J I
then raise intervalError
Page 147 of 709
else if l = u
then l Go Back
Full Screen
Close
Quit
Shrink2: SML ... contd IIT Delhi
else S. Arun-Kumar, CSE
let val m = (l+u) div 2; Title Page
val msqr = m*m Contents
in if msqr <= n JJ II
then shrink (n, m, u) J I
else shrink (n, l, m) Page 148 of 709
end; Go Back
Full Screen
Back to More Algos

Close
Quit
2.2. Algorithm Refinement
1. Recap: More Algorithms
IIT Delhi
2. Recap: Power
3. Recap: Technical completeness S. Arun-Kumar, CSE

Title Page
6. Intsqrt: Reorganized Contents
7. Some More Algorithms

JJ II
8. Combinations: Math
9. Combinations: Details
J I
10. Combinations: SML
11. Perfect Numbers Page 149 of 709
12. Refinement
Go Back
13. Perfect Numbers: SML
Pu
14. l if divisor(k)
Full Screen
15. SML: sum divisors

Close
16. if divisor and ifdivisor
17. SML: Assembly 1 Quit
18. SML: Assembly 2
19. Perfect Numbers . . . contd.
21. SML: Assembly 3 IIT Delhi
22. Perfect Numbers: Run

S. Arun-Kumar, CSE
24. SML: Code variations Title Page
25. SML: Code variations

Contents
27. Summation: Generalizations JJ II
28. Algorithmic Improvements:
29. Algorithmic Variations J I
30. Algorithmic Variations

Page 150 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Recap: More Title Page
Algorithms Contents
• xn for real and integer x

JJ II
J I
• Integer square root Page 151 of 709
Forward
Go Back
Full Screen
Close
Quit
Recap: Power
• xn for real and integer x
IIT Delhi
S. Arun-Kumar, CSE
– Technical Completness Title Page
∗ Undefinedness Contents
∗ Termination JJ II
– More complete definition for real J I
x Page 152 of 709
– Power of an integer Go Back
– ⊥ and exceptions Full Screen
• Integer square root Close
Quit
Recap: Technical
completeness IIT Delhi
S. Arun-Kumar, CSE
Can it be always guaranteed that

Title Page
• x will be real? YES Contents
• n will be integer? YES JJ II
• n will be non-negative? NO J I
• x 6= 0? NO Page 153 of 709
If x = 0 what is 0.00?
Go Back
Full Screen
Close
INFINITE COMPUTATION Quit

Recap: More IIT Delhi
Algorithms S. Arun-Kumar, CSE
• xn for real and integer x Title Page
• Integer square root Contents
JJ II
– Analysis J I
– Algorithmic idea Page 154 of 709
– Algorithm Go Back
– where Full Screen
– and let ...in ...end Close
Quit
Intsqrt: Reorganized
exception negError; IIT Delhi
exception intervalError; S. Arun-Kumar, CSE
fun intsqrt n = Title Page
let fun shrink (n, l, u) = Contents
if l>u orelse JJ II
l*l > n orelse
J I
u*u < n
Page 155 of 709
then raise intervalError

Go Back
else if (l+1)*(l+1) <= n
Full Screen
then shrink (n, l+1, u)
Close
else l;
Quit
Intsqrt: Reorganized IIT Delhi
S. Arun-Kumar, CSE
in if n<0 Title Page
then raise negError Contents
else if n=0 JJ II
then 0 J I
else shrink (n, 0, n) Page 156 of 709
end Go Back
Full Screen
Back
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Title Page
Some More Algorithms Contents
• Combinations JJ II
J I
• Perfect Numbers
Page 157 of 709
Go Back
Full Screen
Close
Quit
Combinations: Math
nC = n!
k (n−k)!k! IIT Delhi
S. Arun-Kumar, CSE
n(n−1)···(n−k+1)
= k!
Title Page
Contents
n(n−1)···(k+1) JJ II
= (n−k)! J I
= n−1Ck−1 + n−1Ck
Page 158 of 709
Go Back
Since we already have the function Full Screen
fact, we may program nCk using Close
any of the above identities. Let’s pro-

gram it using the last one. Quit
Combinations: Details
Given a set of n ≥ 0 elements, find IIT Delhi
the number of subsets of k elements, S. Arun-Kumar, CSE
where 0 ≤ k ≤ n Title Page


 ⊥ if n < 0 or Contents
k < 0 or JJ II




k>n

 J I

nC = 1 if n = 0 or Page 159 of 709
k
k = 0 or

 Go Back


k=n

 Full Screen

 n−1
Ck−1 + n−1Ck otherwise

Close
Quit
Combinations: SML
exception invalid_arg;
fun comb (n, k) = IIT Delhi
if n < 0 orelse S. Arun-Kumar, CSE
k < 0 orelse Title Page
k > n Contents
then raise invalid_arg JJ II
else if n = 0 orelse J I
k = 0 orelse Page 160 of 709
n = k Go Back
then 1
Full Screen
else comb (n-1, k-1) +

Close
comb (n-1, k);

Quit
Back to Some More Algorithms

Perfect Numbers
An integer n > 0 is perfect if it equals IIT Delhi
the sum of all its proper divisors. S. Arun-Kumar, CSE
A divisor k|n is proper if 0 < k < n

Title Page
k|n ⇐⇒ n mod k = 0 Contents
JJ II
perf ect(n) J I
Page 161 of 709

P
⇐⇒ n = {k : 0 < k < n, k|n} Go Back
Full Screen
Pn−1
⇐⇒ n = k=1 if divisor(k) Close
where Quit
IIT Delhi
S. Arun-Kumar, CSE
Refinement Title Page
Contents
1. if divisor(k) needs to be defined JJ II

Pn−1
2. k=1 if divisor(k) needs to be de- J I
fined algorithmically. Page 162 of 709
Go Back
Full Screen
Close
Quit
Perfect Numbers: SML IIT Delhi
exception nonpositive; S. Arun-Kumar, CSE
fun perfect (n) = Title Page
if n <= 0 Contents
then raise nonpositive JJ II
else J I
n = sum_divisors (1, n-1) Page 163 of 709
Go Back
where sum divisors needs to be Full Screen
defined
Close
Quit
Pu
l if divisor(k) IIT Delhi
Pu
k=l if divisor(k) = S. Arun-Kumar, CSE
Title Page



 0 if l > u Contents
 JJ II
 if divisor(l)+ otherwise J I
 n−1 if divisor(k)

 P Page 164 of 709
k=l+1
Go Back
where if divisor(k) needs to be de- Full Screen
fined Close
Quit
SML: sum divisors
From
Pu the algorithmic definition of IIT Delhi
k=l if divisor(k) S. Arun-Kumar, CSE
Title Page
Contents
fun sum_divisors (l, u) =
JJ II
if l > u
then 0 J I
else ifdivisor (l) + Page 165 of 709
sum_divisors (l+1, u) Go Back
Full Screen
where if divisor(k) still needs to be Close
defined Quit
if divisor and ifdivisor
IIT Delhi
k if k|n
if divisor(k) = S. Arun-Kumar, CSE
0 otherwise
Title Page
Contents
fun ifdivisor (k) = JJ II
if n mod k = 0 J I
then k Page 166 of 709
else 0 Go Back
Full Screen
Not technically complete! Close
However . . . Quit
SML: Assembly 1
fun sum_divisors (l, u) = IIT Delhi
if l > u then 0 S. Arun-Kumar, CSE
else
Title Page
let fun ifdivisor (k) =

Contents
if n mod k = 0
JJ II
then k
J I
else 0
Page 167 of 709
in ifdivisor (l) +
Go Back
sum_divisors (l+1, u)
end Full Screen
Close
Clearly k ∈ [l, u] Quit

SML: Assembly 2
exception nonpositive;
fun perfect (n) = IIT Delhi
if n <= 0 S. Arun-Kumar, CSE
then raise nonpositive Title Page
else Contents
let fun sum_divisors (l, u) = JJ II
... J I
in n = sum_divisors (1, n-1) Page 168 of 709
end Go Back
Full Screen
Clearly k ∈ [l, u] = [1, n − 1] whenever Close
n > 0. Quit
Technically complete!
Perfect Numbers
. . . contd. IIT Delhi
S. Arun-Kumar, CSE
Clearly for all k, n/2 < k < n,

Title Page
if divisor(k) = 0.
Contents
bn/2c = n div 2 < n/2 JJ II
J I
Hence
Page 169 of 709
n−1
X n div
X 2 Go Back
if divisor(k) = if divisor(k) Full Screen
k=1 k=1 Close
Quit
Perfect Numbers
. . . contd. IIT Delhi
Hence S. Arun-Kumar, CSE
perf ect(n) Title Page
Contents
Pn−1
⇐⇒ n = k=1 if divisor(k) JJ II
J I
Pn div 2
⇐⇒ n = k−1 if divisor(k) Page 170 of 709
where Go Back
Full Screen
k if k|n
if divisor(k) = Close
0 otherwise
Quit
SML: Assembly 3
S. Arun-Kumar, CSE
if n <= 0 Title Page
then raise nonpositive Contents
else JJ II
let fun sum_divisors (l, u) = J I

...
Page 171 of 709
in n = sum_divisors (1, n div 2)

Go Back
end
Full Screen
Close
Clearly k ∈ [l, u] = [1, n div2] when-

Quit
ever n > 0.
Perfect Numbers: Run IIT Delhi
exception nonpositive S. Arun-Kumar, CSE
val perfect = fn : int -> bool Title Page
val it = () : unit Contents
- perfect ˜8; JJ II
uncaught exception nonpositive J I
raised at: perfect.sml:4.16-4.27

Page 172 of 709
- perfect 5; Go Back
val it = false : bool Full Screen
Close
Quit
Perfect Numbers: Run IIT Delhi
- perfect 6;
S. Arun-Kumar, CSE

Title Page
- perfect 23;
Contents
val it = false : bool

JJ II
- perfect 28;
J I
GC #0.0.0.1.3.88: (1 ms)
Page 173 of 709
- perfect 30; Go Back
Close
Quit
SML: Code variations
if n <= 0 S. Arun-Kumar, CSE
then raise nonpositive

Title Page
else
Contents
let
JJ II
fun ifdivisor (k) = ...;
J I
fun sum_divisors (l, u) = ...
Page 174 of 709
in
Go Back
n=sum_divisors (1, n div 2)
end Full Screen
Close
Technically complete though Quit
ifdivisor, by itself is not!

What about this? IIT Delhi
S. Arun-Kumar, CSE
fun perfect (n) =

Title Page
let
Contents
fun ifdivisor (k) = ...;

JJ II
fun sum_divisors (l, u) = ...
J I
in if n <= 0
Page 175 of 709
then raise nonpositive
else Go Back
n=sum_divisors (1, n div 2) Full Screen
end Close
Quit
What about this? IIT Delhi
exception nonpositive; S. Arun-Kumar, CSE
fun ifdivisor (k) = ...; Title Page
fun sum_divisors (l, u) = ...; Contents
fun perfect (n) = JJ II
if n <= 0 J I
then raise nonpositive Page 176 of 709
else Go Back
n=sum_divisors (1, n div 2) Full Screen
Close
Technically incomplete! Quit

Summation:
Generalizations IIT Delhi
S. Arun-Kumar, CSE
Need a method to compute summa-

Title Page
tions in general.
Contents
For any function f : Z → Z and inte-

JJ II
gers l and u,
 J I
0 if l > u
u

 Page 177 of 709
X 
f (i) = Go Back
 fP(l)+ otherwise
i=l  u
 Full Screen
i=l+1 f (i) Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Algorithmic Title Page
Improvements:
Contents
JJ II
1. perf ect2 J I
2. shrink2 Page 178 of 709
Go Back
Full Screen
Close
Quit
Algorithmic Variations
1. For any k|n, m = n div k is also a IIT Delhi
divisor of n S. Arun-Kumar, CSE
Title Page
2. 1 is a divisor of every positive num-

Contents
ber
√ JJ II
3. For n > 2, b nc < n div 2 J I

Pn div 2
4. Hence k=1 if divisor(k) = Page 179 of 709
√ Go Back
b nc
X Full Screen
1+ if divisor2(k)
Close
k=2
Quit
Algorithmic Variations
IIT Delhi
perf ect(n) S. Arun-Kumar, CSE
Pb√nc Title Page
⇐⇒ n = 1 + k=2 if divisor2(k) Contents
where JJ II
 J I
 k+
Page 180 of 709
if divisor2(k) = (n div k) if k|n
0 otherwise Go Back

Full Screen
Are there any glitches? Is it techni- Close
cally correct and complete?

Quit
2.3. Variations: Algorithms & Code
IIT Delhi
S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 181 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
Recap S. Arun-Kumar, CSE
Title Page
• Combinations Contents
• Perfect Numbers JJ II
• Code Variations J I
• Algorithmic Variations Page 182 of 709
Go Back
forward
Full Screen
Close
Quit
Recap: Combinations IIT Delhi
nC = n! S. Arun-Kumar, CSE
k (n−k)!k! Title Page
Contents
n(n−1)···(n−k+1)
= k! JJ II
J I
n(n−1)···(k+1)
= (n−k)!
Page 183 of 709
Go Back
= n−1Ck−1 + n−1Ck Full Screen
Close
Quit
Combinations 1 IIT Delhi
use "fact.sml";
S. Arun-Kumar, CSE
Title Page
fun comb_wf (n, k) =

Contents
if n < 0 orelse
JJ II
k < 0 orelse
J I
k > n
Page 184 of 709
then raise invalid_arg
else fact (n) div Go Back
(fact(n-k) * fact(k)); Full Screen
Close
Quit
Combinations 2
k > n Contents

k = 0 orelse
Page 185 of 709
n = k
Go Back
then 1
Full Screen
else (* 0<k<n *)
Close
prod (n, n-k+1) div
fact (k) Quit
Combinations 3
k > n Contents

k = 0 orelse
Page 186 of 709
n = k
Go Back
then 1
Full Screen
else (* 0<k<n *)
Close
prod (n, k+1) div
fact (n-k) Quit
Perfect 2
perf ect(n) IIT Delhi
S. Arun-Kumar, CSE
Pb√nc Title Page
⇐⇒ n = 1 + k=2 if divisor2(k)
Contents
where JJ II
if divisor2(k) = J I

 k + m if k|n and k = 6 m Page 187 of 709
k if k|n and k = m Go Back
0 otherwise
 Full Screen
Close
where m = (n div k)
Quit
IIT Delhi
Power 2 S. Arun-Kumar, CSE
power Title Page
The previous inductive definition used Contents
xn = (x × x × · · · × x) ×x
JJ II
| {z } J I
n−1 times
Page 188 of 709
We could associate the product differ- Go Back
ently Full Screen
Close
Quit
A Faster Power
xn = |(x × x ×{z· · · × x)} IIT Delhi
n/2 times
S. Arun-Kumar, CSE
× (x
| ×x× {z· · · × x)} Title Page
n/2 times Contents
when n is even and JJ II
xn = |(x × x × {z· · · × x)}

J I
bn/2c times
Page 189 of 709
× |(x × x ×{z· · · × x)}

Go Back
Full Screen
bn/2c times
× x Close
Quit
when n is odd
Power2: Complete
IIT Delhi
S. Arun-Kumar, CSE
power2(x, n) = Title Page


 1.0/power2(x, n) if n < 0 Contents
 1.0 if n = 0 JJ II
(power2(x, bn/2c)) 2 if even(n) J I



(power2(x, bn/2c))2 × x otherwise
 Page 190 of 709
Go Back
where even(n) ⇐⇒ n mod 2 = 0. Full Screen
Close
Quit
IIT Delhi
Power2: SML S. Arun-Kumar, CSE
fun power2 (x, n) = Title Page
if n < 0 Contents
then 1.0/power2 (x, ˜n) JJ II
else if n = 0 J I
then 1.0 Page 191 of 709
else Go Back
Full Screen
Close
Quit
Power2: SML
let fun even m = IIT Delhi
(m mod 2 = 0);
S. Arun-Kumar, CSE
fun square y = y * y;
Title Page
val pwr_n_by_2 =
Contents
power2 (x, n div 2);

JJ II
val sq_pwr_n_by_2 =
J I
square (pwr_n_by_2)
Page 192 of 709
in if even (n)
then sq_pwr_n_by_2 Go Back
else x * sq_pwr_n_by_2 Full Screen
end Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Computation: Issues Title Page
1. Correctness Contents
JJ II
(a) General correctness
J I
(b) Technical Completeness
Page 193 of 709
(c) Termination
Go Back
Full Screen
Close
Quit
General Correctness IIT Delhi
1. Mathematical correctness should S. Arun-Kumar, CSE
be established for all algorithmic Title Page
variations. Contents
JJ II
2. Program Correctness: Mathemati-
J I
cally developed code should not be
moved around arbitrarily. Page 194 of 709
Go Back
• Code variations should also be

Full Screen
mathematically proven
Close
Quit
Code: Justification IIT Delhi
• How does one justify the correct- S. Arun-Kumar, CSE
ness of Title Page
– this version and Contents
JJ II
– this version?
J I
• Can one correct this version? Page 195 of 709
• But first of all, what is incorrect Go Back
about this version? Full Screen
incorrectness Close
Quit
Recall
• A program is an IIT Delhi
– explicit, S. Arun-Kumar, CSE
– unambiguous and Title Page
– technically complete Contents
JJ II
translation of an algorithm written in
J I
mathematical notation.
Page 196 of 709
• Moreover, mathematical notation is Go Back
more concise than a program. Full Screen
• Hence mathematical notation is Close
easier to analyse and diagnose. Quit

Features: Definition
before Use IIT Delhi
S. Arun-Kumar, CSE
incorrect version
Title Page
Definition of a name before use: Contents
• if divisor(k) is defined first. JJ II
• idivisor(k) uses the name n without J I
defining it. Page 197 of 709
• k has been defined (as an argu- Go Back
ment of if divisor(k)) before being Full Screen
used. Close
Quit
Run ifdivisor IIT Delhi

S. Arun-Kumar, CSE
- fun ifdivisor(k) =
Title Page
= if n mod k = 0
Contents
= then k
JJ II
= else 0
J I
;
Page 198 of 709
stdIn:18.8 Error:
unbound variable Go Back
or constructor: n Full Screen
Close
Quit
Diagnosis: Features of IIT Delhi
programs S. Arun-Kumar, CSE
incorrect version Title Page
• So both sum divisors(l, u) and Contents
perf ect(n) may use if divisor(k). JJ II
J I
• sum divisors(l, u) is defined before
perf ect(n). Page 199 of 709
Go Back
• So perf ect(n) may use both

Full Screen
if divisor(k) and sum divisors(l, u)

Close
Quit
Back to Math
incorrect version
Let IIT Delhi

k if k|n S. Arun-Kumar, CSE
if divisor(k) =
0 otherwise Title Page
Contents
and sum divisors(l, u) = JJ II


0 if l > u J I
if divisor(l)+ otherwise Page 200 of 709
sum divisors(l + 1, u)

Go Back
and perf ect(n) ⇐⇒ Full Screen
Close
n = sum divisors(1, bn/2c) Quit

Incorrectness IIT Delhi
incorrect version S. Arun-Kumar, CSE
• if divisor(k) has a single argument

Title Page
k Contents
JJ II
• But it actually depends upon n too! J I
• But that is not made explicit in its Page 201 of 709
definition. Go Back
Let’s make it explicit! Full Screen
Close
Quit
ifdivisor3
Let IIT Delhi

k if k|n
if divisor3(n, k) = S. Arun-Kumar, CSE
and sum divisors(l, u) = Contents
 JJ II
0 if l > u
J I
if divisor3(n, l)+ otherwise
Page 202 of 709

Go Back
n = sum divisors(1, bn/2c) Close
Quit
Run it! IIT Delhi
Standard ML of New Jersey S. Arun-Kumar, CSE
- fun ifdivisor3 (n, k) Title Page
= = if (n mod k = 0) Contents
= then k JJ II
= else 0; J I
val ifdivisor3 = Page 203 of 709
fn : int * int -> int Go Back
- Full Screen
Close
Quit
Try it!
IIT Delhi
- fun sum_divisors (l, u) =
= if l > u S. Arun-Kumar, CSE
= then 0 Title Page
= else ifdivisor3 (n, l)+ Contents
= sum_divisors (l+1, u); JJ II
stdIn:40.18 Error: unbound J I
variable or constructor: n Page 204 of 709
- Go Back
Full Screen
Now sum divisors also depends on Close
n! Quit
IIT Delhi
S. Arun-Kumar, CSE
Hey! Wait a minute! Title Page
But n was defined in ifdivisor3 (n, k)! Contents
JJ II
So then where is the problem? J I
Let’s ignore it for the moment and Page 205 of 709
come back to it later Go Back
Full Screen
Close
Quit
The n’s
Let IIT Delhi

k if k|n S. Arun-Kumar, CSE
if divisor3(n, k) =
Contents
and sum divisors2(n, l, u) =

JJ II

0 if l > u J I
if divisor3(n, l)+ otherwise Page 206 of 709

Go Back
Close
n = sum divisors2(n, 1, bn/2c) Quit

Scope IIT Delhi
S. Arun-Kumar, CSE
• The scope of a name begins from Title Page
its definition and ends where the Contents
corresponding scope ends JJ II
• Scopes end with definitions of func- J I
tions Page 207 of 709
• Scopes end with the keyword end Go Back
in any let ... in ...end Full Screen
Close
Quit
Scope Rules IIT Delhi
S. Arun-Kumar, CSE
• Scopes may be disjoint Title Page
• Scopes may be nested one Contents
completely within another JJ II
• A scope cannot span two disjoint J I
scopes Page 208 of 709
• Two scopes cannot (partly) overlap Go Back
Full Screen
forward
Close
Quit
2.4. Names, Scopes & Recursion
1. Disjoint Scopes
IIT Delhi
2. Nested Scopes
3. Overlapping Scopes S. Arun-Kumar, CSE
4. Spannning
Title Page
5. Scope & Names
6. Names & References Contents
7. Names & References

JJ II
J I
11. Names & References Page 209 of 709

Go Back
14. Names & References Full Screen

Close
16. Definition of Names
17. Use of Names Quit
18. local...in...end
21. local...in...end IIT Delhi
22. Scope & local

S. Arun-Kumar, CSE
23. Computations: Simple
24. Simple computations Title Page
25. Computations: Composition

Contents
26. Composition: Alternative
27. Compositions: Compare JJ II
28. Compositions: Compare
29. Computations: Composition J I
30. Recursion
Page 210 of 709
31. Recursion: Left
32. Recursion: Right Go Back
Full Screen
Close
Quit
Disjoint Scopes
let
val x = 10; IIT Delhi
fun fun1 y = S. Arun-Kumar, CSE
let
... Title Page
in Contents
... JJ II
end
J I
fun fun2 z =
let Page 211 of 709
...
in
Go Back
in ... Full Screen
end
fun1 (fun2 x)
Close
end Quit
Nested Scopes
let
Title Page
let Contents
val x = 15 JJ II
in J I
x + y Page 212 of 709
end Go Back
in Full Screen
fun1 x Close
end Quit
Overlapping Scopes
let
... Title Page
Contents
... JJ II
J I
Page 213 of 709
... Go Back
in ... Full Screen
fun1 (fun2 x)
Close
end Quit
Spannning
let
... Title Page
Contents
... JJ II
J I
fun fun2 z = Page 214 of 709
... Go Back
in ... Full Screen
fun1 (fun2 x)
Close
end Quit
Scope & Names IIT Delhi
• A name may occur either as being S. Arun-Kumar, CSE
defined or as a use of a previously Title Page
defined name Contents
• The same name may be used to re- JJ II
fer to different objects. J I
• The use of a name refers to the Page 215 of 709
textually most recent definition in Go Back
the innermost enclosing scope Full Screen
diagram Close
Quit
Names & References
let
val x = 10; val z = 5;
IIT Delhi
fun fun1 y =
S. Arun-Kumar, CSE
let Title Page
val x = 15 Contents
JJ II
in
J I
x + y * z
end
Page 216 of 709
in Go Back
Full Screen
fun1 x
Close
end Quit
Back to scope names

Names & References
let
val x = 10; val z = 5;
IIT Delhi
fun fun1 y =
S. Arun-Kumar, CSE
let Title Page
val x = 15 Contents
JJ II
in
J I
x + y * z
end
Page 217 of 709
in Go Back
Full Screen
fun1 x
Close
end Quit
Back to scope names

Names & References
let
val x = 10; val z = 5;
IIT Delhi
fun fun1 y =
S. Arun-Kumar, CSE
let Title Page
val x = 15 Contents
JJ II
in
J I
x + y * z
end
Page 218 of 709
in Go Back
Full Screen
fun1 x
Close
end Quit
Back to scope names

Names & References
let
val x = 10; val z = 5;
IIT Delhi
fun fun1 y =
S. Arun-Kumar, CSE
let Title Page
val x = 15 Contents
JJ II
in
J I
x + y * z
end
Page 219 of 709
in Go Back
Full Screen
fun1 x
Close
end Quit
Back to scope names

Names & References
let
val x = 10; val z = 5;
IIT Delhi
fun fun1 y =
S. Arun-Kumar, CSE
let Title Page
val x = 15 Contents
JJ II
in
J I
x + y * z
end
Page 220 of 709
in Go Back
Full Screen
fun1 x
Close
end Quit
Back to scope names

Names & References
let
val x = 10; val z = 5;
IIT Delhi
fun fun1 y =
S. Arun-Kumar, CSE
let Title Page
val x = 15 Contents
JJ II
in
J I
x + y * z
end
Page 221 of 709
in Go Back
Full Screen
fun1 x
Close
end Quit
Back to scope names

Names & References
let
val x = 10; val z = 5;
IIT Delhi
fun fun1 y =
S. Arun-Kumar, CSE
let Title Page
val x = 15 Contents
JJ II
in
J I
x + y * z
end
Page 222 of 709
in Go Back
Full Screen
fun1 x
Close
end Quit
Back to scope names

Names & References
let
val x = 10; val x = x − 5;
IIT Delhi
fun fun1 y =
let S. Arun-Kumar, CSE
...
in Title Page
... Contents
end JJ II
fun fun2 z = J I
let
... Page 223 of 709
in
... Go Back
end Full Screen
in fun1 (fun2 x) Close
end Quit
Back to scope names

Names & References
let
val x = 10; val x = x − 5;
IIT Delhi
fun fun1 y =
...
in Title Page
... Contents
end JJ II
fun fun2 z = J I
let
... Page 224 of 709
in
... Go Back
end Full Screen
end Quit
Back to scope names

Names & References
let
val x = 10; val x = x − 5;
IIT Delhi
fun fun1 y =
...
in Title Page
... Contents
end JJ II
fun fun2 z = J I
let
... Page 225 of 709
in
... Go Back
end Full Screen
end Quit
Back to scope names

Definition of Names IIT Delhi
S. Arun-Kumar, CSE
Definitions are of the form Title Page
qualifier name . . . = body Contents
• val name = JJ II
• fun name ( argnames ) = J I
• local def initions Page 226 of 709
in def inition Go Back
end Full Screen
Close
Quit
Use of Names IIT Delhi
Names are used in expressions. S. Arun-Kumar, CSE
Expressions may occur Title Page
• by themselves – to be evaluated Contents
• as the body of a definition JJ II
• as the body of a let-expression J I
Page 227 of 709

let def initions
in expression Go Back
end Full Screen
Close
use of local
Quit
local...in...end
perfect
IIT Delhi
local
S. Arun-Kumar, CSE
exception invalidArg;
Title Page
Contents
fun ifdivisor3 (n, k) =
JJ II
if n <= 0 orelse
J I
k <= 0 orelse
n < k Page 228 of 709
then raise invalidArg Go Back
else if n mod k = 0 Full Screen
then k Close
else 0; Quit
local...in...end
perfect
fun sum_div2 (n, l, u) = IIT Delhi
if n <= 0 orelse S. Arun-Kumar, CSE
l <= 0 orelse Title Page
l > n orelse Contents
u <= 0 orelse JJ II
u > n J I
then raise invalidArg Page 229 of 709
else if l > u Go Back
then 0 Full Screen
else ifdivisor3 (n, l) Close
+ sum_div2 (n, l+1, u) Quit

local...in...end
perfect
IIT Delhi
in S. Arun-Kumar, CSE
fun perfect n =
Title Page
if n <= 0
Contents
then raise invalidArg
JJ II
else
J I
let
val nby2 = n div 2 Page 230 of 709
in Go Back
n = sum_div2 (n, 1, nby2) Full Screen
end Close
end Quit
local...in...end
perfect
- use "perfect2.sml"; S. Arun-Kumar, CSE
[opening perfect2.sml] Title Page
GC #0.0.0.0.1.10: (1 ms) Contents
val perfect = fn : int -> bool JJ II
val it = () : unit J I
- perfect 28; Page 231 of 709
val it = true : bool Go Back
- perfect 6; Full Screen

Close
- perfect 8128;
Quit

Scope & local
local
fun fun1 y = IIT Delhi
...
S. Arun-Kumar, CSE
Title Page
Contents
fun fun2 z = ... JJ II

fun1
J I
Page 232 of 709
in Go Back
fun fun3 x = ...

fun2 ... Full Screen
fun1 ... Close
end Quit
Computations: Simple IIT Delhi
For most simple expressions it is S. Arun-Kumar, CSE
• left to right, and Title Page
• top to bottom Contents
JJ II
except when
J I
• presence of brackets
Page 233 of 709
• precedence of operators Go Back
determine otherwise. Full Screen
Hence Close
Quit
IIT Delhi
Simple computations S. Arun-Kumar, CSE
Title Page
4 + 6 − (4 + 6) div 2 Contents
= 10 − (4 + 6) div 2 JJ II
= 10 − 10 div 2 J I
= 10 − 5 Page 234 of 709
= 5 Go Back
Full Screen
Close
Quit
Computations: Composition
f (x) = x2 + 1 IIT Delhi
S. Arun-Kumar, CSE
g(x) = 3 ∗ x + 2 Title Page
Then for any value a = 4 Contents
f (g(a)) JJ II
= f (3 ∗ 4 + 2) J I
= f (14) Page 235 of 709
= 142 + 1 Go Back
= 196 + 1 Full Screen
= 197 Close
Quit
Composition:
Alternative
f (x) = x2 + 1
IIT Delhi
S. Arun-Kumar, CSE
g(x) = 3 ∗ x + 2 Title Page
Why not Contents
JJ II
f (g(a))
g(4)2 + 1
J I
=
(3 ∗ 4 + 2)2 + 1
Page 236 of 709
=
(12 + 2)2 + 1
Go Back
=
142 + 1
Full Screen
=
Close
= 196 + 1
= 197 Quit
Compositions:
Compare
IIT Delhi
S. Arun-Kumar, CSE
Title Page
f (g(a)) f (g(a))
= f (3 ∗ 4 + 2) = g(4)2 + 1
Contents
= (3 ∗ 4 + 2)2 + 1
JJ II
= f (14)
= (12 + 2)2 + 1
J I
=
= 142 + 1 = 142 + 1
Page 237 of 709
Go Back
= 196 + 1 = 196 + 1
Full Screen
= 197 = 197
Close
Quit
IIT Delhi
Compositions: S. Arun-Kumar, CSE
Compare Title Page
Contents
Question 1: Which is more correct? JJ II
Why? J I
Question 2: Which is easier to imple- Page 238 of 709
ment?
Question 3: Which is more efficient? Go Back
Full Screen
Close
Quit
IIT Delhi
Computations: S. Arun-Kumar, CSE
Composition Title Page
A computation of f (g(a)) proceeds Contents
thus: JJ II
J I
• g(a) is evaluated to some value, say
Page 239 of 709
b
Go Back
• f (b) is next evaluated Full Screen
Close
Quit
IIT Delhi
Recursion S. Arun-Kumar, CSE
Title Page

1 if n = 0 Contents
f actL(n) =
f actL(n − 1) ∗ n otherwise JJ II
J I
1 if n = 0
f actR(n) = Page 240 of 709
n ∗ f actR(n − 1) otherwise
Go Back
Full Screen
Close
Quit
Recursion: Left IIT Delhi
S. Arun-Kumar, CSE
f actL(4) Title Page
= (f actL(4 − 1) ∗ 4) Contents
= (f actL(3) ∗ 4) JJ II
= ((f actL(3 − 1) ∗ 3) ∗ 4) J I
= ((f actL(2) ∗ 3) ∗ 4) Page 241 of 709
= (((f actL(2 − 1) ∗ 2) ∗ 3) ∗ 4) Go Back
= ···
Full Screen
Close
Quit
Recursion: Right IIT Delhi
S. Arun-Kumar, CSE
f actR(4) Title Page
= (4 ∗ f actR(4 − 1)) Contents
= (4 ∗ f actR(3)) JJ II
= (4 ∗ (3 ∗ f actR(3 − 1))) J I
= (4 ∗ (3 ∗ f actR(2))) Page 242 of 709
= (4 ∗ (3 ∗ (2 ∗ f actR(2 − 1)))) Go Back
= ··· Full Screen
Close
Quit
3. Introducing Reals
3.1. Floating Point IIT Delhi
1. So Far-1: Computing
S. Arun-Kumar, CSE
2. So Far-2: Algorithms & Programs
3. So far-3: Top-down Design Title Page
4. So Far-4: Algorithms to Programs

Contents
5. So far-5: Caveats
6. So Far-6: Algorithmic Variations JJ II
7. So Far-7: Computations
J I
8. Floating Point
9. Real Operations Page 243 of 709
10. Real Arithmetic

Go Back
11. Numerical Methods
12. Errors
Full Screen
13. Errors
14. Infinite Series Close
15. Truncation Errors

Quit
16. Equation Solving
17. Root Finding-1
18. Root Finding-2
19. Root Finding-3 IIT Delhi
20. Root Finding-4

S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 244 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
So Far-1: Computing S. Arun-Kumar, CSE
• The general nature of computation Title Page
• The notion of primitives, composi- Contents
tion & induction JJ II
J I
• The notion of an algorithm
Page 245 of 709
• The digital computer & program- Go Back
ming language Full Screen
Close
Quit
So Far-2: Algorithms & IIT Delhi
Programs S. Arun-Kumar, CSE
• Algorithms: Finite mathematical Title Page
processes Contents
JJ II
• Programs: Precise, unambiguous
J I
explications of algorithms
Page 246 of 709
• Standard ML: Its primitives Go Back
• Writing technically complete speci- Full Screen
fications Close
Quit
So far-3: Top-down
Design
integer Square Root IIT Delhi
• Begin with the function you need to S. Arun-Kumar, CSE
design Title Page
Contents
• Write a
JJ II
– small compact technically com- J I
plete definition of the function Page 247 of 709
– perhaps using other functions Go Back
that have not yet been defined Full Screen
• Each function in turn is defined in a Close
top-down manner Quit
Perfect Numbers
So Far-4: Algorithms to IIT Delhi
Programs S. Arun-Kumar, CSE
• Perform top development till you re-

Title Page
quire only the available primitives Contents
JJ II
• Directly translate the algorithm into
J I
a Program
Page 248 of 709
• Use scope rules to localize or gen- Go Back
eralize Full Screen
SML code for perfect

Close
Quit
So far-5: Caveats
IIT Delhi
S. Arun-Kumar, CSE
• Don’t arbitrarily vary code from your

Title Page
algorithmic development
Contents
– It might work or JJ II
– It might not work J I
– unless properly justified Page 249 of 709
• May destroy technical complete- Go Back
ness Full Screen
• May create scope violations. Close
Quit
So Far-6: Algorithmic IIT Delhi
Variations S. Arun-Kumar, CSE
Title Page
Algorithmic Variations Contents
• Are safe if developed from first prin- JJ II
ciples. Thus ensuring their J I
– mathematical correctness Page 250 of 709
– technical completeness Go Back
– termination properties Full Screen
Close
Quit
So Far-7: IIT Delhi
Computations S. Arun-Kumar, CSE
Title Page
• Work within the notion of mathe-

Contents
matical equality
JJ II
– Simple expressions J I
– Composition of functions Page 251 of 709
– Recursive computations Go Back
• But are generally irreversible Full Screen
Close
Quit
Floating Point
• Each real number 3E11 is repre- IIT Delhi
sented by a pair of integers S. Arun-Kumar, CSE
1. Mantissa: 3 or 30 or 300 or . . . Title Page
2. Exponent: the power of 10 which Contents
the mantissa has to be multiplied JJ II
by J I
• What is displayed is not necessarily Page 252 of 709
the same as the internal represen- Go Back
tation. Full Screen
• There is no unique representation Close
of a real number Quit

Real Operations IIT Delhi
Depending upon the operations in- S. Arun-Kumar, CSE
volved Title Page
• Each real number is first converted Contents
into a suitable representation JJ II
J I
• The operation is performed
Page 253 of 709
• The result is converted into a suit- Go Back
able representation for display. Full Screen
skip to Numerical methods

Close
Quit
Real Arithmetic
• for addition and subtraction the two IIT Delhi
numbers should have the same ex- S. Arun-Kumar, CSE
ponent for ease of integer opera- Title Page
tions to be performed Contents
• the conversion may involve loss of JJ II
precision J I
• for multiplication and division the Page 254 of 709
exponents may have to be adjusted Go Back
so as not to cause an integer over- Full Screen
flow or underflow. Close
back Quit
Numerical Methods IIT Delhi
S. Arun-Kumar, CSE
• Finite (limited) precision

Title Page
• Accuracy depends upon available Contents
precision JJ II
• Whereas integer arithmetic is exact, J I
real arithmetic is not. Page 255 of 709
• Numerical solutions are a Go Back
finite approximation of the result Full Screen
Close
Quit
IIT Delhi
Errors S. Arun-Kumar, CSE
• Hence an estimate of the error is Title Page
necessary. Contents
• If a is the “correct” value and a∗ is JJ II
the computed value, J I
absolute error = a∗ − a
Page 256 of 709
a ∗−a Go Back
relative error = a
Full Screen
Close
Quit
Errors IIT Delhi
Errors in floating point computations S. Arun-Kumar, CSE
are mainly due Title Page
Contents
finite precision Round-off errors

JJ II
fnite process It is impossible to com- J I
pute the value of a (convergent) in- Page 257 of 709
finite series because computations Go Back
are themselves finite processes. In- Full Screen
finite series
Close
Quit
Infinite Series
cannot be computed to ∞
IIT Delhi
∞
X xm
ex = S. Arun-Kumar, CSE
m! Title Page
m=0
Contents
∞
X (−1)mx2m JJ II
cos x =
(2m)! J I
m=0 Page 258 of 709
∞ Go Back
X (−1)mx2m+1
sin x = Full Screen
(2m + 1)!
m=0 Close
Quit
Truncation
Truncation Errors
and hopefully it is good enough to re-
strict it to appropriate values of n IIT Delhi
n
X xm S. Arun-Kumar, CSE
ex =
m! Title Page
m=0 Contents
n JJ II
X (−1)mx2m
cos x = J I
(2m)!
m=0 Page 259 of 709
Go Back
n
X (−1)mx2m+1 Full Screen
sin x =
(2m + 1)! Close
m=0
Quit
IIT Delhi
S. Arun-Kumar, CSE
Equation Solving Title Page
• The fifth most basic operation Contents
• Root finding: A particular form of JJ II
equation solving J I
Page 260 of 709
f (x) = 0 Go Back
Full Screen
Close
Quit
Root Finding-1
f(b)
IIT Delhi
S. Arun-Kumar, CSE
Title Page
Contents
x0
JJ II
a
J I
b
Page 261 of 709
Go Back
f(a) Full Screen
Close
Quit
Root Finding-2
f(b)
IIT Delhi
S. Arun-Kumar, CSE
Title Page
Contents
x0
JJ II
a
J I
b
Page 262 of 709
Go Back
f(a) Full Screen
Close
Quit
Root Finding-3
f(b)
IIT Delhi
S. Arun-Kumar, CSE
Title Page
Contents
x0
JJ II
a
J I
b
Page 263 of 709
Go Back
f(a) Full Screen
Close
Quit
Root Finding-4
Rather steep isn’t it?
IIT Delhi
f(b) S. Arun-Kumar, CSE
Title Page
Contents
x0 JJ II
a J I
b Page 264 of 709
Go Back
f(a)
Full Screen
Close
ε
Quit
3.2. Root Finding, Composition and Recursion
1. Root Finding: Newton’s Method
IIT Delhi
3. Root Finding: Newton’s Method S. Arun-Kumar, CSE

Title Page
6. Root Finding: Newton’s Method Contents
7. Newton’s Method: Basis

JJ II
9. Newton’ Method: Algorithm
J I
10. What can go wrong!-1
11. What can go wrong!-2 Page 265 of 709

Go Back
14. What can go wrong!-4 Full Screen
15. Real Computations & Induction: 1

Close
17. What’s it good for? 1 Quit
18. What’s it good for? 2
19. newton: Computation
20. Generalized Composition
21. Two Computations of h(1) IIT Delhi
22. Two Computations of h(−1)

S. Arun-Kumar, CSE
23. Recursive Computations
24. Recursion: Left Title Page
25. Recursion: Right

Contents
26. Recursion: Nonlinear
27. Some Practical Questions JJ II
28. Some Practical Questions
J I
Page 266 of 709
Go Back
Full Screen
Close
Quit
Root Finding: Newton’s IIT Delhi
Method S. Arun-Kumar, CSE
Consider a function f (x) Title Page
Contents
• smooth and continuously differen-

JJ II
tiable over [a, b]
J I
• with a non-zero derivative f 0(x) ev- Page 267 of 709
erywhere in [a, b] Go Back
• the signs of f (a) and f (b) are differ- Full Screen
ent Close
Quit
Root Finding: Newton’s
Method IIT Delhi
S. Arun-Kumar, CSE
f(b)
Title Page
Contents
JJ II
J I
Page 268 of 709
a Go Back
b
Full Screen
f(a)
Close
Quit
Method IIT Delhi
S. Arun-Kumar, CSE
f(b)
Title Page
Contents
JJ II
J I
Page 269 of 709
a xi
Go Back
b
Full Screen
f(a)
Close
Quit
Method IIT Delhi
S. Arun-Kumar, CSE
f(b)
Title Page
Contents
JJ II
J I
Page 270 of 709
xi+1 xi
a αi Go Back
b
Full Screen
f(a)
Close
Quit
Method IIT Delhi
S. Arun-Kumar, CSE
f(b)
Title Page
Contents
JJ II
J I
Page 271 of 709
xi+1 xi
a αi Go Back
b
Full Screen
f(a)
Close
Quit
Method IIT Delhi
S. Arun-Kumar, CSE
f(b)
Title Page
Contents
JJ II
J I
Page 272 of 709
xi+1 xi
a αi Go Back
xi+2 b
Full Screen
f(a)
Close
Quit
Newton’s Method:
Basis IIT Delhi
f (x )
= f 0(xi) = x −xi
S. Arun-Kumar, CSE
tan αi
i i+1 Title Page
whence Contents
f (x )
xi+1 = xi − f 0(xi ) JJ II
i
J I
Starting from an initial value x0 ∈ [a, b],
if the sequence f (xi) converges to 0 Page 273 of 709
i.e Go Back
Full Screen
f (x0), f (x1), f (x2), · · · → 0 Close
Quit
Newton’s Method:
Basis IIT Delhi
i.e. lim |f (xn)| = 0

S. Arun-Kumar, CSE
n→∞ Title Page
Contents
i.e.∀ε > 0 : ∃N ≥ 0 : ∀n > N : JJ II
J I
|f (xn)| < ε Page 274 of 709
then the sequence Go Back
Full Screen
x 0 , x1 , x2 , · · ·
Close
converges to a root of f . Quit

Newton’ Method:
Algorithm
Select a small enough ε > 0 and x0. IIT Delhi
Then S. Arun-Kumar, CSE
newton(f, f 0, a, b, ε, xi) = Title Page

if |f (xi)| < ε
Contents
xi
newton(f, f 0, a, b, ε, xi+1) otherwise JJ II
J I
x0 ∈ [a, b] Go Back
and Full Screen
f (xi) Close
xi+1 = xi − 0 ∈ [a, b]
f (xi) Quit
What can go wrong!-1
Oscillations! IIT Delhi
S. Arun-Kumar, CSE
f(b)
Title Page
Contents
JJ II
J I
Page 276 of 709
xi+1 xi
a αi Go Back
xi+2 b
Full Screen
f(a)
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
What can go wrong!-2 Title Page
An intermediate point may lie outside Contents
[a, b]! The function may not satisfy all JJ II
the assumptions outside [a, b]. There J I
are then no guarantees about the be- Page 277 of 709
haviour of the function. Go Back
Full Screen
Close
Quit
Interval bounds error! IIT Delhi
S. Arun-Kumar, CSE
f(b)
Title Page
Contents
JJ II
J I
Page 278 of 709
xi+1 xi
a αi Go Back
b
Full Screen
f(a) xi+2
Close
Quit
The function may be too steep
IIT Delhi
f(b) S. Arun-Kumar, CSE
Title Page
Contents
x0
JJ II
a
J I
b
Page 279 of 709
f(a)
Go Back
Full Screen
ε Close
for the available precision. Quit

Or too shallow! IIT Delhi
S. Arun-Kumar, CSE
Title Page
Contents
x0 JJ II
f(b)
a J I
b
f(a) Page 280 of 709
Go Back
Full Screen
Close
Quit
Real Computations &
Induction: 1 IIT Delhi
Newton’s method (when it does work!) S. Arun-Kumar, CSE
computes a sequence Title Page
x 0 , x 1 , x 2 , . . . xn Contents
JJ II
of essentially discrete values such J I
that even if the sequence is not totally Page 281 of 709
ordered, there is some discrete con- Go Back
vergence measure viz.

Full Screen
|f (xi) − 0| Close
which is well-founded. Quit

Real Computations &
Induction: 2 IIT Delhi
S. Arun-Kumar, CSE
That is, for some decreasing se- Title Page
quence of integers ki ≥ 0, Contents
k0 > k 1 > k 2 > · · · > k n = 0 JJ II
J I
we have Page 282 of 709
kiε ≤ |f (xi) − 0| < (ki + 1)ε Go Back
Full Screen
and therefore inductive on integer
multiples of ε Close
Quit
What’s it good for? 1
√
n
Finding the positive n-th root c of a IIT Delhi
c > 0 and n > 1 amounts to solving S. Arun-Kumar, CSE
the equation Title Page
xn = c Contents
which is equivalent to finding the root JJ II
of f (x) with J I
f (x) = xn − c
Page 283 of 709
f 0(x) = nxn−1
Go Back
√ Full Screen
with [a, b] = [0, c] or [0, c] and an ap- Close
propriately chosen ε. Quit

What’s it good for? 2
Finding roots of polynomials.
IIT Delhi
n
ci xi
X S. Arun-Kumar, CSE
f (x) =
Title Page
i=0
Contents
n JJ II
icixi−1
X
f 0(x) = J I
i=1 Page 284 of 709
and Go Back
• an appropriately chosen ε, Full Screen
• an appropriately chosen [a, b] with Close
one of the limits possibly being c0. Quit

newton : Computation IIT Delhi
S. Arun-Kumar, CSE
newton(f, f 0, a, b, ε, x0) Title Page
newton(f, f 0, a, b, ε, x1) Contents
newton(f, f 0, a, b, ε, x2) JJ II
newton(f, f 0, a, b, ε, x3) J I
.. .. ..
Page 285 of 709
newton(f, f 0, a, b, ε, xn) Go Back
xn Full Screen
Close
Quit
Generalized
Composition IIT Delhi
S. Arun-Kumar, CSE
Computations
Title Page
h(x) = f (x, g(x)) Contents
where JJ II
0 if x < 0
f (x, y) = J I
y otherwise Page 286 of 709
Go Back
0 if x = 0
g(x) = Full Screen
g(x − 1) + 1 otherwise Close
Quit
Two Computations of IIT Delhi
S. Arun-Kumar, CSE
h(1)
Title Page
h(1) | h(1) Contents
f (1, g(1)) | f (1, g(1)) JJ II
g(1) | f (1, (g(0) + 1)) J I
g(0) + 1 | f (1, (0 + 1) Page 287 of 709
0+1 | f (1, 1) Go Back
1 | 1 Full Screen
Close
Quit
Two Computations of IIT Delhi
S. Arun-Kumar, CSE
h(−1)
Title Page
h(−1) | h(−1) Contents
f (−1, g(−1)) | f (−1, g(−1)) JJ II
0 | f (−1, (g(−2) + 1)) J I
DONE! | f (−1, ((g(−3) + 1) + 1 Page 288 of 709
| ... Go Back
| FOREVER! Full Screen
Close
Quit
Recursive IIT Delhi
Computations S. Arun-Kumar, CSE
Title Page
• Newton’s method Contents
JJ II
• Factorial
J I
– f actL Page 289 of 709
– f actR Go Back
skip to nonlinear recursion Full Screen

skip to Recursion Revisited
Close
Quit
Recursion: Left IIT Delhi
S. Arun-Kumar, CSE
f actL(4) Title Page
(f actL(4 − 1) ∗ 4) Contents
(f actL(3) ∗ 4) JJ II
((f actL(3 − 1) ∗ 3) ∗ 4) J I
((f actL(2) ∗ 3) ∗ 4) Page 290 of 709
(((f actL(2 − 1) ∗ 2) ∗ 3) ∗ 4) Go Back
···
Full Screen
Close
Quit
Recursion: Right IIT Delhi
S. Arun-Kumar, CSE
f actR(4) Title Page
(4 ∗ f actR(4 − 1)) Contents
(4 ∗ f actR(3)) JJ II
(4 ∗ (3 ∗ f actR(3 − 1))) J I
(4 ∗ (3 ∗ f actR(2))) Page 291 of 709
(4 ∗ (3 ∗ (2 ∗ f actR(2 − 1)))) Go Back
··· Full Screen
Close
Quit
Recursion: Nonlinear IIT Delhi
Fibonacci
S. Arun-Kumar, CSE
f ib(5) Title Page
f ib(4) + f ib(3) Contents
(f ib(3) + f ib(2)) + f ib(3) JJ II
((f ib(2) + f ib(1)) + f ib(2)) + f ib(3) J I
(((f ib(1) + f ib(0)) + f ib(1)) + f ib(2)) + f ib(3)

Page 292 of 709
(((1 + f ib(0)) + f ib(1)) + f ib(2)) + f ib(3) Go Back
··· Full Screen
contd ... Close
Quit
Some Practical
Questions
• What is the essential difference be- IIT Delhi
tween the computations of newton S. Arun-Kumar, CSE
and the two factorial programs? Title Page
Answer: Constant space vs. Linear Contents
space JJ II
• What is the essential similarity be- J I
tween the computations of f actL Page 293 of 709
and f actR? Answer Go Back
• Why can’t we calculate beyond Full Screen
f ib(43) using the definition Fi- Close
bonacci, on ccsun50 or a P-IV? Quit
Answer
Some Practical IIT Delhi
Questions S. Arun-Kumar, CSE
Title Page
• What does a computation of Fi- Contents
bonacci look like? JJ II
• What is the essential difference J I
between the computations of Fi- Page 294 of 709
bonacci and newton or f actL or Go Back
f actR? Full Screen
Close
Quit
4. Correctness, Termination & Complexity
4.1. Termination and Space Complexity
1. Recursion Revisited
IIT Delhi
2. Linear Recursion: Waxing
3. Recursion: Waning S. Arun-Kumar, CSE
4. Nonlinear Recursions
Title Page
5. Fibonacci: contd
6. Recursion: Waxing & Waning

Contents
7. Unfolding Recursion
8. Non-termination JJ II
9. Termination
J I
10. Proofs of termination
11. Proofs of termination: Induction Page 295 of 709
12. Proof of termination: Factorial

Go Back
14. Fibonacci: Termination Full Screen
15. GCD computations

Close
16. Well-foundedness: GCD
17. Well-foundedness
Quit
18. Induction is Well-founded
20. Where it doesn’t work
21. Well-foundedness is inductive IIT Delhi
22. Well-foundedness is inductive

S. Arun-Kumar, CSE
23. GCD: Well-foundedness
24. Newton: Well-foundedness
Title Page
26. Example: Zero Contents
27. Questions
JJ II
28. The Collatz Problem
29. Questions
J I
30. Space Complexity
31. Newton & Euclid: Absolute Page 296 of 709
32. Newton & Euclid: Relative

Go Back
33. Deriving space requirements
34. GCD: Space
Full Screen
35. Factorial: Space
36. Fibonacci: Space Close
37. Fibonacci: Space

Quit
IIT Delhi
S. Arun-Kumar, CSE
Recursion Revisited Title Page
• Linear recursions Contents
JJ II
– Waxing
J I
– Waning
Page 297 of 709
• Non-linear recursion Go Back
Full Screen
Close
Quit
Linear Recursion: IIT Delhi
Waxing S. Arun-Kumar, CSE
Title Page
f actL(4) Contents
(f actL(3) ∗ 4) JJ II
((f actL(2) ∗ 3) ∗ 4) J I
(((f actL(1) ∗ 2) ∗ 3) ∗ 4) Page 298 of 709
((((f actL(0) ∗ 1) ∗ 2) ∗ 3) ∗ 4) Go Back
contrast with newton Full Screen
Close
Quit
IIT Delhi
Recursion: Waning S. Arun-Kumar, CSE
Title Page
((((1 ∗ 1) ∗ 2) ∗ 3) ∗ 4) Contents
(((1 ∗ 2) ∗ 3) ∗ 4) JJ II
((2 ∗ 3) ∗ 4) J I
(6 ∗ 4)
Page 299 of 709
24
Go Back
contrast with newton

Full Screen
Close
Quit
IIT Delhi
Nonlinear Recursions S. Arun-Kumar, CSE
Title Page
Fibonacci
Contents
• Each computation of f ib has its own JJ II
waxing and waning J I
• There is still an “envelope” which Page 300 of 709
shows waxing and waning. Go Back
Full Screen
Close
Quit
IIT Delhi
Fibonacci: contd
S. Arun-Kumar, CSE
Title Page
(((1 + 1) + f ib(1)) + f ib(2)) + f ib(3) Contents
(2 + f ib(1)) + f ib(2)) + f ib(3) JJ II
((2 + 1) + f ib(2)) + f ib(3) J I
··· Page 301 of 709
Go Back
Full Screen
Close
Quit
Recursion: Waxing &
Waning
• Waning: Occurs when an expres- IIT Delhi
sion is simplified without requiring S. Arun-Kumar, CSE
replacement of names by defini- Title Page
tions. Contents
JJ II
• Waxing: Occurs when a name is
J I
replaced by its definition.
Page 302 of 709
– name by value replacements

Go Back
– occurs in generalized composi- Full Screen
tion but just once if it is not recur- Close
sively defined
Quit
– Unfolding recursion
IIT Delhi
S. Arun-Kumar, CSE
Unfolding Recursion Title Page
• may occur several times (terminat- Contents
ing), or JJ II
J I
• even an infinite number of times
Page 303 of 709
leading to nontermination
Go Back
Full Screen
Close
Quit
Non-termination IIT Delhi
Algorithm S. Arun-Kumar, CSE
• Simple expressions never lead to Title Page
nontermination Contents
JJ II
• (Generalized) composition never
leads to nontermination J I
Page 304 of 709
• Recursion may lead to non-

Go Back
termination or infinite computations,

Full Screen
unless proved otherwise

Close
Quit
Termination IIT Delhi
S. Arun-Kumar, CSE
Since recursion may lead to nontermi- Title Page
nation Contents
• Termination needs to be proved for JJ II
recursive definitions, and J I
• for expressions and definitions that Page 305 of 709
use recursively defined names as Go Back
components. Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Proofs of termination Title Page
A recursive definition guarantees ter- Contents
mination JJ II
J I
• if it is inductive, or
Page 306 of 709
• it is well-founded Go Back
Full Screen
Close
Quit
Proofs of termination:
Induction IIT Delhi
S. Arun-Kumar, CSE
A recursive definition guarantees ter-

Title Page
mination
Contents
• if it is inductive, JJ II
Examples: J I
– Factorial Page 307 of 709
– Fibonacci Go Back
• it is well-founded, though not obvi- Full Screen
ously inductive Close
Quit
Proof of termination:
Factorial IIT Delhi
S. Arun-Kumar, CSE
Factorial
Title Page
Consider f actL defined only for non- Contents
negative integers. We prove that it is JJ II
an algorithm i.e. that it terminates J I
Basis : For n = 0, f actL(n) = 1 which Page 308 of 709
is not a recursive definition. Hence Go Back
it does indeed terminate in a single Full Screen
step. Close
Quit
Proof of termination: IIT Delhi
Induction hypothesis . For some n > Title Page
0, ∀k : 0 ≤ k ≤ n : f actL(k) termi- Contents
nates in ∝ k steps. JJ II
Induction step . Then f actL(n + 1) = J I
f actL(n) ∗ (n + 1) is guaranteed to Page 309 of 709
terminate in ∝ (n + 1) steps, since Go Back
f actL(n) does so in ∝ n steps. Full Screen
back Close
Quit
Fibonacci: Termination
Fibonacci
The proof is similar to that of f actL. IIT Delhi
S. Arun-Kumar, CSE
Basis For n = 0 or n = 1 f ib(n) = 1.

Title Page
Induction hypothesis For some n > 0, Contents
∀k : 0 ≤ k ≤ n : f ib(k) terminates in JJ II
∝ f (k) steps
J I
Induction Step Then since each of Page 310 of 709
f ib(n) and f ib(n − 1) is guaranteed

Go Back
to terminate in ∝ f (n) and ∝ f (n −

Full Screen
1) steps f ib(n+1) = f ib(n)+f ib(n−

Close
1) is also guaranteed to terminate in

f (n + 1) ∝ f (n) + f (n − 1) steps.
Quit
IIT Delhi
GCD computations S. Arun-Kumar, CSE
Euclidean GCD Title Page
gcd(12, 64) Contents
gcd(64, 12) JJ II
gcd(12, 4) J I
gcd(4, 0) Page 311 of 709
4 Go Back
Full Screen
Close
Quit
Well-foundedness:
GCD IIT Delhi
Euclidean GCD S. Arun-Kumar, CSE
For x, y > 0, 0 ≤ x mod y < y. Hence Title Page
the sequence of remainders obtained Contents
is JJ II
J I
• a sequence of non-negative integers,
and Page 312 of 709
Go Back
• is strictly decreasing
Full Screen
r1 > r2 > · · · > rn−1 > rn = 0 Close
Quit
Well-foundedness
A definition is well-founded if it is pos- IIT Delhi
sible to define a measure (i.e. a func- S. Arun-Kumar, CSE
tion w of its arguments) called the Title Page
well-founded function such that Contents
JJ II
1. the well-founded function takes only
J I
non-negative integer values
Page 313 of 709
2. with each successive recursive call Go Back
the value of the well-founded func- Full Screen
tion is guaranteed to decrease by at

Close
least 1.
Quit
Induction is IIT Delhi
Well-founded S. Arun-Kumar, CSE
The well-founded function usually is a Title Page
measure of the number of computa- Contents
tion steps that the algorithm will take JJ II
to terminate J I
• Factorial w(n) ∝ n Page 314 of 709
Go Back
• Fibonacci w(n) ∝ f (n)

Full Screen
Then Close
Quit
Induction is IIT Delhi
Well-founded S. Arun-Kumar, CSE
Title Page
• w(n) is always non-negative if f actL Contents
and f ib are defined only for non- JJ II
negative integers J I
• The argument to f actL and f ib in Page 315 of 709
each recursive unfolding is strictly Go Back
decreasing. Full Screen
Close
Quit
Where it doesn’t work IIT Delhi
Such proofs do not work for S. Arun-Kumar, CSE
• f act arbitrarily extended to include Title Page
negative integers. (since w(n) no Contents
longer strictly non-negative) JJ II
• f act(n) = f act(n+1) div (n+1), even J I
if n is non-negative (since w(n) is no Page 316 of 709
longer decreasing) Go Back
since the function is no longer well- Full Screen
founded. Close
Quit
IIT Delhi
Well-foundedness is S. Arun-Kumar, CSE
inductive Title Page
But the induction variable is Contents
• hidden or JJ II
J I
• too complex to worry about, or
Page 317 of 709
• it serves no useful purpose for the Go Back
algorithm, except as a counter. Full Screen
Close
Quit
Well-foundedness is
inductive IIT Delhi
Given any well-founded function w(~x) S. Arun-Kumar, CSE
whose values form a decreasing se- Title Page
quence in some algorithm Contents
y0 > y1 > · · · > yn−1 > yn ≥ 0 JJ II
J I
it is possible to put this sequence Page 318 of 709
in 1-1 correspondence with the set

Go Back
{0, . . . , n} via a function ind such that

Full Screen
ind(w(~x)) = n − i Close
Quit
IIT Delhi
GCD: S. Arun-Kumar, CSE
Well-foundedness Title Page
Contents
GCD
JJ II
Well-founded function for gcd J I
w(a, b) = b Page 319 of 709
Go Back
Full Screen
Close
Quit
Newton:
Well-foundedness IIT Delhi
Newton’s Method
S. Arun-Kumar, CSE
Convergence condition Title Page
f (x0), f (x1), f (x2), · · · → 0 Contents
JJ II
Compute the discrete value sequence J I
x 0 , x 1 , x 2 , . . . xn Page 320 of 709
Go Back
such that
Full Screen
k0 > k 1 > k 2 > · · · > k n = 0 Close
where Quit
Newton: IIT Delhi
Well-foundedness S. Arun-Kumar, CSE
Newton’s Method Title Page
Contents
kiε ≤ |f (xi) − 0| < (ki + 1)ε

JJ II
and therefore inductive on integer J I
multiples of ε Hence Page 321 of 709
|f (x)| Go Back
w(x) = b c
ε Full Screen
Close
Quit
Example: Zero IIT Delhi
A peculiar way to define the zero func- S. Arun-Kumar, CSE
tion Title Page
zero(x) = Contents
 JJ II
 zero(x + 1.0) if x ≤ −1.0
J I
0.0 if −1.0 < x < 1.0
Page 322 of 709
zero(x − 1.0) if x ≥ 1.0


Go Back
w(x) = d|x|e is the well-founded func- Full Screen
tion Close
Quit
Questions IIT Delhi
S. Arun-Kumar, CSE
Q: Is it always possible to find a well- Title Page
founded function for each algorithm? Contents
A: Unfortunately not! However if we JJ II
can’t then we cannot call it an algo- J I
rithm!. But if we can then we are guar- Page 323 of 709
anteed that the algorithm will termi-

nate. Go Back
The Collatz Problem Full Screen
Close
Quit
The Collatz Problem IIT Delhi
Does the following algorithm termi- S. Arun-Kumar, CSE
nate? Title Page
collatz(m) = Contents
 JJ II
1 if m ≤ 1 J I
collatz(m div 2) if m is even Page 324 of 709
collatz(3 ∗ m + 1) otherwise

Go Back
Unproven Claim. collatz(m) 1 for all Full Screen
m.
Close
Quit
Questions
Q: what other uses can well-founded IIT Delhi
functions be put to? S. Arun-Kumar, CSE
A: They can be used to estimate Title Page
the complexity of your algorithm in Contents
order of magnitude terms. JJ II
Space Complexity : The amount of J I
memory space required, as a func- Page 325 of 709
tion of the input Go Back
Time Complexity : The amount of time Full Screen
(number of computation steps) as a Close
function of the input Quit

IIT Delhi
Space Complexity S. Arun-Kumar, CSE
Title Page
What is the space complexity of

Contents
• Newton’s method JJ II
• Euclidean GCD J I
• Factorial Page 326 of 709
• Fibonacci Go Back
Full Screen
Close
Quit
Newton & Euclid:
Absolute
Newton’s Method IIT Delhi
Computation
S. Arun-Kumar, CSE
Newton’s method (wherever and Title Page
whenever it works well) requires Contents
space to compute JJ II
• the value of f at each point xi J I
• the value of f 0 at each point xi Page 327 of 709
• the value of xi+1 from the above Go Back
Their absolute space requirements Full Screen
could be different. But . . . Close
Euclidean GCD Quit
Computation
Newton & Euclid: IIT Delhi
Relative S. Arun-Kumar, CSE
Newton’s Method
Title Page
Computation
GCD and Newton’s method (wherever Contents
and whenever it works well) require JJ II
the same amount of space for each J I
recursive unfolding since each fresh Page 328 of 709
unfolding can reuse the space used

by the previous one. Go Back
Euclidean GCD Full Screen
Computation Close
Quit
Deriving space
IIT Delhi
S. Arun-Kumar, CSE
requirements Title Page
We may use the algorithm itself to de- Contents
rive the space required as follows: JJ II
J I
Assume that memory proportional to
calculating and outputting the answer Page 329 of 709
is a unit. Then space as a function of Go Back
the input is given by Full Screen
Close
Quit
GCD: Space IIT Delhi

1 if b = 0 S. Arun-Kumar, CSE
Sgcd(a,b) =
Sgcd(b,a mod b) otherwise Title Page
Contents
This implies (from well-foundedness)

JJ II
that the entire computation ends with
J I
the space of a unit.
Page 330 of 709
Sgcd(a,b) ∝ 1 Go Back
Full Screen
A similar analysis and result holds for

newton Close
Quit
Factorial: Space
f actL
IIT Delhi
1 if n = 0
Sf actL(n) = S. Arun-Kumar, CSE
Sf actL(n−1)+1 otherwise Title Page
The 1 is for output and the +1 is be- Contents
cause one needs to store space pro- JJ II
portional to remembering “multiply by J I
n”. Page 331 of 709
Go Back
Sf actL(n) ∝ n. Full Screen
Close
A similar analysis and result holds for Quit

f actR.
IIT Delhi
S. Arun-Kumar, CSE
Fibonacci: Space Title Page
Fibonacci Contents
JJ II
1 if n ≤ 1
Sf ib(n) = J I
Sf ib(n−1) + Sf ib(n−2) if n > 1 Page 332 of 709
Go Back
Full Screen
Close
Quit
Fibonacci: Space
Fibonacci
It is easy to see prove by induction IIT Delhi
that for n > 1, S. Arun-Kumar, CSE
Sf ib(n−1) < Sf ib(n) ≤ 2Sf ib(n−1) Title Page
Contents
That is, as the value of n increases JJ II
by 1 the space requirement approxi- J I
mately doubles. Further, it is easy to Page 333 of 709
show by induction that Go Back
2n−2 < Sf ib(n) ≤ 2n−1 Full Screen
Close
Quit
4.2. Efficiency Measures and Speed-ups
1. Recapitulation
IIT Delhi
2. Recapitulation
3. Time & Space Complexity S. Arun-Kumar, CSE
4. isqrt: Space
Title Page
5. Time Complexity
6. isqrt: Time Complexity Contents
7. isqrt2: Time
JJ II
8. shrink vs. shrink2: Times
9. Factorial: Time Complexity
J I
10. Fibonacci: Time Complexity
11. Comparative Complexity Page 334 of 709
12. Comparisons
Go Back
13. Comparisons
14. Efficiency Measures: Time Full Screen
15. Efficiency Measures: Space

Close
16. Speeding Up: 1
17. Speeding Up: 2 Quit
18. Factoring out calculations
19. Tail Recursion: 1
21. Factorial: Tail Recursion IIT Delhi
22. Factorial: Tail Recursion

S. Arun-Kumar, CSE
23. A Computation
24. Factorial: Issues Title Page
25. Fibonacci: Tail Recursion

Contents
27. fibTR: SML JJ II
28. State in Tail Recursion
29. Invariance J I
Page 335 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Recapitulation Title Page
• Recursion & nontermination Contents
• Termination & well-foundedness JJ II
J I
• Well-foundedness proofs
Page 336 of 709
• Well-foundedness & Complexity Go Back
Full Screen
Close
Quit
Recapitulation
• Recursion & nontermination IIT Delhi
S. Arun-Kumar, CSE
• Termination & well-foundedness

Title Page
• Well-foundedness proofs Contents
– By induction JJ II
– well-founded functions J I
– By well-founded functions Page 337 of 709
– induction as well-foundedness Go Back
– Well-foundedness as induction Full Screen
Close
• Well-foundedness & Complexity

Quit
Time & Space
Complexity IIT Delhi
Questions
S. Arun-Kumar, CSE
An order of magnitude estimate of Title Page
the time or space (memory) required Contents
(in terms of some large computation JJ II

steps).
J I
• Newton & Euclid’s GCD Page 338 of 709
• Deriving space requirements Go Back
– Integer Sqrt Full Screen
– Factorial Close
Quit
– Fibonacci
isqrt : Space
Integer Sqrt shrink IIT Delhi
Sisqrt(n) = Sshrink(n,0,n) for large n. S. Arun-Kumar, CSE
Sshrink(n,l,u) = Title Page
 Contents
1
 if l = u JJ II
Sshrink(n,l+1,u) if l < u . . . J I
shrink(n,l,u−1) if l < u . . .
S

Page 339 of 709
Assuming 1 unit of space for output. Go Back
By induction on |[l, u]| Full Screen
Close
Sisqrt(n) = Sshrink(n,0,n) ∝ 1
Quit
Time Complexity IIT Delhi
S. Arun-Kumar, CSE
As in the case of space we may use Title Page
the algorithm itself to derive the time Contents
complexity. JJ II
• Integer sqrt J I
• Factorial Page 340 of 709
• Fibonacci Go Back
Full Screen
forward
Close
Quit
isqrt: Time Complexity
Integer Sqrt
shrink IIT Delhi
S. Arun-Kumar, CSE
Assume condition-checking (along
with +1 or −1) takes a unit of time. Title Page
Then Tshrink(n,l,u) = Contents
JJ II

0
 if l = u J I
1 + Tshrink(n,l+1,u) if l < u . . . Page 341 of 709
shrink(n,l,u−1) if l < u . . .
1+T
 Go Back
Full Screen
Then Tshrink(n,l,u) ∝ |[l, u]| − 1 and

Close
Tisqrt(n) = Tshrink(n,0,n) ∝ n Quit

isqrt2: Time
shrink
Assume condition-checking (along

with (l + u) div 2) takes a unit of time. IIT Delhi
Then Tshrink2(n,l,u) = S. Arun-Kumar, CSE
Title Page

0
 if u ≤ l ≤ u Contents
1 + Tshrink2(n,m,u) if m2 ≤ n . . . JJ II
1+T

if m 2>n
shrink2(n,l,u−1) J I
Page 342 of 709
If 2k−1 ≤ |[l, u]| − 1 < 2k then the al- Go Back
gorithm terminates in at most k steps. Full Screen
Since k = dlog2 |[l, u]| − 1e, Close
Tshrink2(n,l,u) ∝ dlog2 |[l, u]| − 1e Quit

shrink vs. shrink2 : Times
shrink shrink2
IIT Delhi
1. The time units are different,

S. Arun-Kumar, CSE
2. But they differ by a constant factor Title Page
at most. Contents
3. So clearly, for large n, shrink2 is JJ II
faster than shrink J I
4. But for small n, it depends on the Page 343 of 709
constant factor. Go Back
5. Implicitly assume that the actual Full Screen
unit of time includes the time re- Close
quired to unfold the recursion. Quit

Factorial: Time
S. Arun-Kumar, CSE
f actL
Title Page
Here we assume multiplication takes Contents
unit time. JJ II

0 if n = 0 J I
Tf actL(n) =
Tf actL(n−1)+1 otherwise Page 344 of 709
Go Back
Then
Full Screen
Tf actL(n) ∝ n Close
Quit
Fibonacci: Time
Fibonacci S. Arun-Kumar, CSE
Assuming addition and condition-

Title Page
checking together take a unit of time, Contents
we have JJ II
J I
0 if n ≤ 1
Tf ib(n) = Page 345 of 709
Tf ib(n−1) + Tf ib(n−2) if n > 1

Go Back
It follows that Full Screen
2n−2 < Tf ib(n) ≤ 2n−1 Close
Quit
IIT Delhi
Comparative S. Arun-Kumar, CSE
Contents
Algorithm Space Time

JJ II
isqrt(n) O(1) O(n)
J I
isqrt2(n) O(1) O(log2 n)
Page 346 of 709
f actL(n) O(n) O(n)
f ib(n) O(2n) O(2n) Go Back
Full Screen
Close
Quit
Comparisons IIT Delhi
For smaller values S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 347 of 709
Go Back
Full Screen
Close
Quit
Comparisons
IIT Delhi
For large values S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 348 of 709
Go Back
Full Screen
Close
Quit
Efficiency Measures: IIT Delhi
Time S. Arun-Kumar, CSE
Title Page
An algorithm for a problem is Contents
asymptotically faster or asymptotically

JJ II
more time-efficient than another for
J I
the same problem if its time complex-
Page 349 of 709
ity is bounded by a function that has

Go Back
a slower growth rate as a function of
the value of its arguments. Full Screen
Close
Quit
IIT Delhi
Efficiency Measures: S. Arun-Kumar, CSE
Space Title Page
Contents
Similarly an algorithm is JJ II
asymptotically more space efficient J I
than another if its space complexity
is bounded by a function that has a Page 350 of 709
slower growth rate. Go Back
Full Screen
Close
Quit
IIT Delhi
Speeding Up: 1 S. Arun-Kumar, CSE
Title Page
Q: Can fibonacci be speeded up or
made more space efficient? Contents
JJ II
A: Perhaps by studying the nature of J I
the function e.g. isqrt2 vs. isqrt
Page 351 of 709
and attempting more efficient algo-

Go Back
rithmic variations.
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Speeding Up: 2 Title Page
Q: Are there general methods of

Contents
JJ II
speeding up or saving space?
J I
A: Take inspiration from gcd, newton,
Page 352 of 709
shrink
Go Back
Full Screen
Close
Quit
Factoring out
calculations IIT Delhi
S. Arun-Kumar, CSE
gcd(a0, b0) Title Page
compute a1, b1 Contents
gcd(a1, b1) JJ II
compute a2, b2 J I
gcd(a2, b2) Page 353 of 709
··· Go Back
gcd(an, bn) Full Screen
an Close
Quit
IIT Delhi
Tail Recursion: 1 S. Arun-Kumar, CSE
• Factor out calculations and remem-

Title Page
ber only those values that are re- Contents
quired for the next recursive call. JJ II
J I
• Create a vector of state variables
Page 354 of 709
and include them as arguments of

Go Back
the function
Full Screen
Close
Quit
Tail Recursion: 2 IIT Delhi
S. Arun-Kumar, CSE
• Try to reorder the computation us- Title Page
ing the state variables so as to get Contents
the next state completely defined. JJ II
• Redefine the function entirely in J I
terms of the state variables so that Page 355 of 709
the recursive call is the outermost Go Back
operation. Full Screen
Close
Quit
Factorial: Tail
Recursion
f actL Waxing f actL Waning IIT Delhi
• The recursive call precedes the S. Arun-Kumar, CSE
multiplication operation. Change it! Title Page
• Define a state variable p which con- Contents
tains the product of all the values JJ II
that one must remember J I
Page 356 of 709

• Reorder the computation so that the
computation of p is performed be-
Go Back
fore the recursive call. Full Screen
Close
• For that redefine the function in

Quit
terms of p.
Factorial: Tail
Recursion IIT Delhi

⊥ if n < 0 Title Page
f actL2(n) = 1 if n = 0 Contents
f actL tr(n, 1) otherwise JJ II


J I
f actL tr(n, p) =
Go Back

p if n = 0 Full Screen
f actL tr(n − 1, np) otherwise Close
Quit
A Computation IIT Delhi
S. Arun-Kumar, CSE
f actL2(4)
Title Page
f actL tr(4, 1)
Contents
f actL tr(3, 4)
JJ II
f actL tr(2, 12)
J I
f actL tr(1, 24)
f actL tr(0, 24) Page 358 of 709
24 Go Back
Full Screen
Reminiscent of gcd and newton! Close
Quit
Factorial: Issues
• Correctness: Prove (by induction
on n) that for all n ≥ 0, f actL2(n) = IIT Delhi
n!. S. Arun-Kumar, CSE
• termination: Prove by induction on Title Page
n that every computation of f actL2 Contents
terminates. JJ II
J I
• Space complexity: Prove that
Sf actL2(n) = O(1) (as against Page 359 of 709
Sf actL(n) ∝ n). Go Back
Full Screen
• Time complexity: Prove that Close
Tf actL2(n) = O(n)
Quit
Complexity table
Fibonacci: Tail
Recursion IIT Delhi
S. Arun-Kumar, CSE
• Remove duplicate computations by Title Page
defining appropriate state variables Contents
• Let a and b be the consecutive fi- JJ II
bonacci numbers f ib(m − 2) and J I
f ib(m − 1) required for the compu- Page 360 of 709
tation of f ib(m). Go Back
• The state consists of the variables Full Screen
n, a, b, m. Close
Quit
Fibonacci: Tail
Recursion IIT Delhi
f ibT R(n) = S. Arun-Kumar, CSE
 Title Page
⊥ if n < 0
if 0 ≤ n ≤ 1
Contents
1
JJ II
f ib iter(n, 1, 1, 1) otherwise

J I
f ib iter(n, a, b, m) = Go Back

b if m ≥ n Full Screen
f ib iter(n, b, a + b, m + 1) otherwise Close
Quit
fibTR: SML
local IIT Delhi
fun fib_iter (n, a, b, m) = S. Arun-Kumar, CSE
(* fib (m) = b ,fib (m-1) = a *) Title Page
if m >= n then b Contents
else fib_iter (n, b, a+b, m+1);JJ II
in J I
fun fibTR (n) = Page 362 of 709
if n < 0 then raise negativeArgumen Go Back
else if (n <= 1) then 1 Full Screen
else fib_iter (n, 1, 1, 1) Close
end; Quit
State in Tail Recursion
IIT Delhi
S. Arun-Kumar, CSE
• The variables that make up the Title Page
state bear a definite relation to each Contents
other. JJ II
• INVARIANCE. That relationship be- J I
tween the state variables does not Page 363 of 709
change throughout the computation Go Back
of the function. Full Screen
Close
Quit
Invariance IIT Delhi
• The invariant property of a tail- S. Arun-Kumar, CSE
recursive function must hold Title Page
Initially when it is first invoked, and Contents
Continues to hold before every suc- JJ II
cessive invocation J I
• The invariant property character- Page 364 of 709
izes the entire computation and the Go Back
algorithm and is crucial to the proof Full Screen
of correctness Close
Quit
4.3. Invariance & Correctness
1. Recap
IIT Delhi
2. Recursion Transformation
3. Tail Recursion: Examples S. Arun-Kumar, CSE
4. Comparisons
Title Page
5. Transformation Issues
6. Correctness Issues 1 Contents
7. Correctness Issues 2
JJ II
8. Correctness Theorem
9. Invariants & Correctness 1
J I
11. Invariance Lemma: f actL tr Page 365 of 709
12. Invariance: Example

Go Back
14. Proof Full Screen
15. Invariance Lemma: f ib iter

Close
16. Proof
17. Correctness: Fibonacci Quit
18. Variants & Invariants
20. More Invariants
21. Fast Powering 1 IIT Delhi
22. Fast Powering 2

S. Arun-Kumar, CSE
23. Root Finding: Bisection
24. Advantage Bisection Title Page
Contents
JJ II
J I
Page 366 of 709
Go Back
Full Screen
Close
Quit
Recap IIT Delhi
• Asymptotic Complexity: S. Arun-Kumar, CSE
Title Page
– Space
Contents
– Time JJ II
• Comparative Complexity J I
• Comparisons: Page 367 of 709
– Small inputs Go Back
– Large inputs Full Screen
Close
Quit
Recursion
Transformation IIT Delhi
S. Arun-Kumar, CSE
• To achieve constant space and lin- Title Page
ear time, if possible Contents
• Speeding up using tail recursion JJ II
– Factor out calculations J I
– Reorder the computations with Page 368 of 709
state variables Go Back
– Recursion as the outermost oper- Full Screen
ation Close
Quit
IIT Delhi
Tail Recursion: S. Arun-Kumar, CSE
Examples Title Page
Contents
• Factorial vs. Factorial: JJ II
f actL vs. f actL2 vs. J I
• Fibonacci vs. Fibonacci: Page 369 of 709
f ib vs. f ibT R Go Back
Full Screen
Close
Quit
Comparisons
IIT Delhi
S. Arun-Kumar, CSE
Algorithm Space Time Title Page
isqrt(n) O(1) O(n) Contents
isqrt2(n) O(1) O(log2 n) JJ II
f actL(n) O(n) O(n) J I
f actL2(n) O(1) O(n) Page 370 of 709
f ib(n) O(2n) O(2n) Go Back
f ibT R(n) O(1) O(n) Full Screen
Close
Quit
Transformation Issues IIT Delhi
S. Arun-Kumar, CSE
• Correctness: Prove that the new al- Title Page
gorithm computes the same func-

Contents
tion as the original simple algorithm

JJ II
• Termination: Prove by induction on J I
n that every computation is finite. Page 371 of 709
• Space complexity: Compute it. Go Back
• Time complexity: Compute it. Full Screen
Close
Quit
Correctness Issues 1 IIT Delhi
S. Arun-Kumar, CSE
• Absolute correctness: For any func- Title Page
tion f , that an algorithm A that Contents
claims to implement it, prove that JJ II

f (~x) = A(~x) J I
for all argument values ~x for which Page 372 of 709
f is defined. Go Back
• Transformation correctness: Full Screen
Close
Quit
Correctness Issues 2 IIT Delhi
• Absolute correctness: S. Arun-Kumar, CSE
• Transformation correctness: For Title Page
any algorithm A and a transformed Contents
algorithm B prove that JJ II
A(~x) = B(~x) J I
for all argument values ~x for which Page 373 of 709
A is defined. Then B is absolutely Go Back
correct provided A is absolutely cor- Full Screen
rect. Close
Quit
Correctness Theorem IIT Delhi
S. Arun-Kumar, CSE
Invariant properties f actL2

Title Page
Theorem 8 For all n ≥ 0, Contents
f actL2(n) = n!
JJ II
Proof: For n = 0, it is clear that J I

f actL2(0) = 1 = 0!. For n > 0,
f actL2(n) = f actL tr(n, 1). The proof Page 374 of 709
is done provided we can show that Go Back
f actL tr(n, 1) = n!. Full Screen
Close
Quit
Invariants & IIT Delhi
Correctness 1 S. Arun-Kumar, CSE
Title Page
Contents
• To prove the absolute or trans-

JJ II
formation correctness of a tail-
J I
recursion transformation usually re-
quires an invariant property to be
Page 375 of 709
proven about the tail-recursive func- Go Back
tion. Full Screen
Close
Quit
Invariants &
Correctness 2 IIT Delhi
S. Arun-Kumar, CSE

Title Page
• This allows the independent proof Contents
of the properties of the tail- JJ II
recursive function without reference J I
to the function that uses it. Page 376 of 709
• It reflects the design of the al- Go Back
gorithm and its division into sub- Full Screen
problems. Close
Quit
IIT Delhi
Invariance Lemma:
f actL tr
S. Arun-Kumar, CSE
Title Page
Invariant properties f actL2 Contents
Lemma 9 For all n ≥ 0 and p JJ II
f actL tr(n, p) = (n!)p J I
Proof: By induction on n. Page 377 of 709
Back to theorem Go Back
Full Screen
Close
Quit
Invariance: Example IIT Delhi
f actL2
S. Arun-Kumar, CSE
f actL tr(4, 7) Title Page
f actL tr(3, 28) Contents
f actL tr(2, 84) JJ II
f actL tr(1, 168) J I
f actL tr(0, 168) Page 378 of 709
168 Go Back
Contrast with a f actL2(4) computa- Full Screen
tion Close
Quit
Invariance: Example IIT Delhi
f actL2 S. Arun-Kumar, CSE
So what exactly is invariant? Title Page
f actL tr(4, 7) 168 = 4! × 7 Contents
f actL tr(3, 28) 168 = 3! × 28 JJ II
f actL tr(2, 84) 168 = 2! × 84 J I
f actL tr(1, 168) 168 = 1! × 168 Page 379 of 709
f actL tr(0, 168) 168 = 0! × 168 Go Back
168 Full Screen
Close
Quit
Proof
Basis For n = 0, f actL tr(0, p) = p = IIT Delhi
(0!)p.
S. Arun-Kumar, CSE
Induction hypothesis (IH) For all k, Title Page
0 < k ≤ n, Contents
f actL tr(k, p) = p = (k!)p JJ II
Induction Step J I
f actL tr(n + 1, p) Page 380 of 709
= f actL tr(n, (n + 1)p) Go Back
= (n!)(n + 1)p (IH) Full Screen
= (n + 1)!p Close
Quit
Back to lemma
Invariance Lemma:
f ib iter
IIT Delhi
S. Arun-Kumar, CSE
f ib iter F Title Page
Lemma 10 For all n > 1, a, b, m : 1 ≤ Contents
m ≤ n, if a = F(m − 1) and b = F(m), JJ II
then J I
INV :f ib iter(n, a, b, m) = F(n) Page 381 of 709
. Go Back
Proof: By induction on k = n − m Full Screen
Close
Quit
Proof
Basis For k = 0, n = m, it follows that
f ib iter(n, a, b, m) = F(n) IIT Delhi
S. Arun-Kumar, CSE
Induction hypothesis (IH) For all n >

1 and 1 ≤ m ≤ n, with n − m ≤ k,
Title Page
INV holds Contents
JJ II
Induction Step Let 1 ≤ m < n such
that n − m = k + 1, F(m) = b and J I
F(m − 1) = a. Then F(m + 1) = a + b Page 382 of 709
and Go Back
Full Screen
f ib iter(n, a, b, m)
Close
= f ib iter(n, b, a + b, m + 1)
= F(n) (IH) Quit
Correctness: Fibonacci IIT Delhi
S. Arun-Kumar, CSE
f ibT R Fibonacci
Title Page
Theorem 11 For all n ≥ 0, Contents
f ibT R(n) = F(n). JJ II
Proof: For 0 ≤ n ≤ 1, it holds J I

trivially. For n > 1, f ibT R(n) =
Page 383 of 709
f ib iter(n, 1, 1, 1) = F(n), by the invari-
ance lemma, with m = 1, a = 1 = Go Back
F(m − 1) and b = 1 = F(m). Full Screen
Close
Quit
Variants & Invariants IIT Delhi
S. Arun-Kumar, CSE
f actL2
Title Page
f actL3(n) = Contents

⊥ if n < 0 JJ II
1 if n = 0 J I
f actL tr2(n, 1, 1) else

 Page 384 of 709
Go Back
where
Full Screen
Close
Quit
Variants & Invariants IIT Delhi
S. Arun-Kumar, CSE
f actL2
Title Page
f actL tr2(n, p, m) = Contents

p if n = m JJ II
f actL tr2(n, (m + 1)p, m + 1) else J I
Page 385 of 709
f actL tr2(n, p, m) = (m!)p Go Back
for all 1 ≤ m ≤ n. Full Screen
Close
Quit
More Invariants IIT Delhi
• shrink For all n > 0, l, u, if S. Arun-Kumar, CSE
[l, u] ⊆ [0, n], Title Page
Contents
√
l ≤ b nc ≤ u JJ II
J I
• shrink2
Page 386 of 709
For all n > 0, l, u, if [l, u] ⊆ [0, n],

Go Back
√
m = b(l + u)/2c and l ≤ b nc ≤ u Full Screen
Close
Quit
IIT Delhi
Fast Powering 1 S. Arun-Kumar, CSE
power2 Title Page
power3(x, n) = Contents
JJ II

 1.0/power3(x, −n) if n < 0
J I
1.0 if n = 0
Page 387 of 709
powerT R(x, n, 1) else

Go Back
where Full Screen
Close
Quit
Fast Powering 2 IIT Delhi
S. Arun-Kumar, CSE
power2
Title Page
powerT R(x, n, p) =
Contents

p if n = 0 JJ II
powerT R(x2, n div 2, p) if even(n) J I
powerT R(x2, n div 2, xp) otherwise


Page 388 of 709
where even(n) ⇐⇒ n mod 2 = 0. Go Back
powerT R(x, n, p) = xnp Full Screen
Close
Quit
Root Finding: Bisection IIT Delhi
Newton’s Method Algorithm S. Arun-Kumar, CSE
Select a small enough ε > 0 and Title Page
x0. Then if sgn(f (a)) 6= sgn(f (b)), Contents
bisect(f, a, b, ε) = JJ II

c if |f (c)| < ε J I
bisect(f, c, b, ε) if sgn(f (c)) 6= sgn(f (b)) Page 389 of 709
bisect(f, a, c, ε) otherwise
 Go Back
Full Screen
where c = (a + b)/2 Close
Quit
Advantage Bisection
More robust than Newton’s method IIT Delhi
• Requires continuity and change of S. Arun-Kumar, CSE
sign Title Page
• Does not require differentiability Contents
JJ II
• Could change the condition suitably J I
to take care of very shallow curves
Page 390 of 709
• Oscillations could occur only if the Go Back
function is too steep. Full Screen
• An intermediate point can never go Close
outside the interval. Quit

5. Compound Data
5.1. Tuples, Lists & the Generation of Primes
1. Recap: Tail Recursion
IIT Delhi
2. Examples: Invariants
3. Tuples S. Arun-Kumar, CSE
4. Lists
Title Page
5. New Lists
6. List Operations
Contents
7. List Operations: cons
8. Generating Primes upto n JJ II
9. More Properties
J I
10. Composites
11. Odd Primes Page 391 of 709
12. primesU pto(n)

Go Back
13. generateF rom(P, m, n, k)
14. generateF rom Full Screen
15. primeW RT (m, P )

Close
17. primeW RT
Quit
18. Density of Primes
19. The Prime Number Theorem
21. Complexity IIT Delhi
22. Diagnosis
S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 392 of 709
Go Back
Full Screen
Close
Quit
Recap: Tail Recursion IIT Delhi
S. Arun-Kumar, CSE
• Asymptotic Complexity:
Title Page
Time Linear Contents
Space Constant JJ II
• Correctness: Capture the algorithm J I
through Page 393 of 709
Invariant Invariance Lemma Go Back
Bound function Proof by induction Full Screen
Close
Quit
Examples: Invariants IIT Delhi
f actL tr2 S. Arun-Kumar, CSE
shrink & shrink2 Title Page
Contents
√
l ≤ b nc ≤ u JJ II
√ J I
m = b(l + u)/2c and l ≤ b nc ≤ u
Page 394 of 709
Fast Powering Go Back
Full Screen
powerT R(x, n, p) = xnp Close
Quit
Tuples: Formation
Simplest form of compound data:
Cartesian products. IIT Delhi
S. Arun-Kumar, CSE
• Each element of a cartesian prod-

Title Page
uct is a tuple
Contents
• Tuples may be constructed as we JJ II
do in mathematics, simply by en- J I

closing the elements (separated by Page 395 of 709
commas) in a pair of parentheses.

Go Back
- val a = (2, 3.0, false); Full Screen
val a = (2,3.0,false) : Close
int * real * bool Quit

Tuples: Decomposition IIT Delhi
• Individual components of a tuple S. Arun-Kumar, CSE
may be taken out too. Title Page
- #1 a; Contents
- #2 (a); J I
val it = 3.0 : real Page 396 of 709
- #3 a; Go Back
Close
Quit
Tuples: divmod
Standard ML of New Jersey, ... IIT Delhi
- fun divmod (a, b) = S. Arun-Kumar, CSE
(a div b, a mod b); Title Page
val divmod = Contents
fn : int * int -> int * int

JJ II
- val dm = divmod (24,9);
J I
val dm = (2,6) : int * int
Page 397 of 709
- #1 dm;
Go Back
val it = 2 : int
Full Screen
- #2 dm;
Close
val it = 6 : int
Quit
Constructors &
Destructors
Every way of constructing compound IIT Delhi
data from simpler data elements has S. Arun-Kumar, CSE
Constructors : Operators which con- Title Page
struct compound data from simpler Contents
ones (for tuples it is simply (, , and JJ II
)). J I
Destructors : Operators which allow Page 398 of 709
us to extract the individual compo- Go Back
nents of a compound data item (for Full Screen
tuples they are #1, #2 ... depend- Close
ing upon how many components it Quit
consists of).
Tuples: Identity IIT Delhi
Every tuple that has been broken up S. Arun-Kumar, CSE
into its components using its destruc- Title Page
tors can be put together back again Contents
using its constructors. JJ II
Given a tuple a ∈ A1 × A2 × . . . × An, J I
we have Page 399 of 709
Go Back
a = (#1 a, #2 a, . . . , #n a) Full Screen
Close
Quit
Lists IIT Delhi
An α list represents a sequence of el- S. Arun-Kumar, CSE
ements of a given type α. Title Page
Given a (nonempty) list Contents
• A list is ordered JJ II
• There may be more than one occur- J I
rence of an element in the list Page 400 of 709
• only the first element (called the Go Back
head) of the list is immediately ac- Full Screen
cessible. Close
Quit
New Lists
Given a (nonempty) list L, IIT Delhi
• A new list M may be created from S. Arun-Kumar, CSE
an existing list L by the tl operation. Title Page
• New elements can be added (by the Contents
operation cons) to an existing list, JJ II
one at a time to create new lists. J I
• the last element that was added be- Page 401 of 709
comes the head of the new list. Go Back
• Two lists are equal only if they have Full Screen
the same elements in the same or- Close
der Quit
List Operations
• The empty list: nil or [] IIT Delhi
S. Arun-Kumar, CSE
• Nonempty lists: Given a nonempty

Title Page
list L
Contents
L = [1, 2, 3, 4]
JJ II
head : hd : αList → α J I
Page 402 of 709

hd(L) = 1
Go Back
tail : tl : αList → αList

Full Screen
tl(L) = [2, 3, 4] Close
Quit
List Operations: cons
IIT Delhi
S. Arun-Kumar, CSE
• L = [1, 2, 3, 4] Title Page
cons : cons : α × αList → αList Contents
JJ II
cons(0, nil) = [0] J I
cons(0, L) = 0 :: L = [0, 1, 2, 3, 4] Page 403 of 709
1 :: (0 :: L) = [1, 0, 1, 2, 3, 4] Go Back
back to lists Recap Full Screen
Close
Quit
Polynomial Evaluation
Evaluating a polynomial
IIT Delhi
n
aixi
X S. Arun-Kumar, CSE
p(x) =
Title Page
i=0
Contents
given JJ II
• its coefficients as a list [an, . . . , a0] J I
of values from the highest degree Page 404 of 709
term to the constant a0. Go Back
• a value for the variable x. Full Screen
Assume an empty list of coefficients Close
yields a value 0. Quit

Naive Solution IIT Delhi
poly0(L, x) = S. Arun-Kumar, CSE
Title Page
0 if L = nil
hxn + poly0(T, x) if L = h :: T
Contents
JJ II
where n = |L| − 1. J I
fun poly0 ([], x) = 0.0 Page 405 of 709
| poly0 ((h::T), x) = Go Back
h*power(x, n)+poly0(T, x) Full Screen
Close
Quit
Complexity of poly0 IIT Delhi
Space. O(n) to store both the list and S. Arun-Kumar, CSE
the intermediate computations. Title Page
Additions. O(n) additions. Contents
JJ II
Multiplications. n(n − 1)/2 by the sim-
J I
plest powering algorithm.
Page 406 of 709
Multiplications. O(log2(n) + O(log2(n − Go Back
1) + · · · + O(log2(1)) ≤ O(n log2(n)) Full Screen
by the fast powering algorithm. Close
Quit
Arden’s Rule
Factor out the multiplications to get IIT Delhi
S. Arun-Kumar, CSE
p(x) = (...((anx+an−1)x+an−2)x+...)x+a0
Title Page
and define a tail-recursive function Contents
which requires only n multiplications. JJ II
poly1(L, x) = poly T R(0, L, x) J I
poly T R(p, L, x) = Go Back

p if L = nil Full Screen
poly T R(px + h, T, x) if L = h :: T Close
Quit
poly1 in SML
local
fun poly_TR (p, [], x) = p IIT Delhi
| poly_TR (p, (h::T), x) = S. Arun-Kumar, CSE
poly_TR (p*x + h, T, x) Title Page
in Contents
fun poly L x = JJ II
poly_TR (0.0, L, x) J I
end; Page 408 of 709
Go Back
Question 1. What is the right theorem Full Screen
to prove that poly T R is the right gen- Close
eralization for the problem?

Quit
Ans.
poly1 in SML
local IIT Delhi
fun poly_TR (p, [], x) = p S. Arun-Kumar, CSE
| poly_TR (p, (h::T), x) = Title Page
poly_TR (p*x + h, T, x) Contents
in JJ II
fun poly L x = J I
poly_TR (0.0, L, x) Page 409 of 709
end; Go Back
Full Screen
n
X
Ans. poly T R(p, L, x) = pxn+1 + aixi Close
i=0
Quit
Reverse Input
IIT Delhi
S. Arun-Kumar, CSE
Supposing the coefficients were given Title Page
in reverse order [a0, . . . , an]. Revers- Contents
ing this list will be an extra O(n) time JJ II
and space. Though the asymptotic J I
complexity does not change much, it Page 410 of 709
is more interesting to work directly Go Back
with the given list. Full Screen
Close
Quit
IIT Delhi
revpoly0 S. Arun-Kumar, CSE
revpoly0(L, x) = Title Page

0 if L = nil Contents
h + x × revpoly0(T, x) if L = h :: T JJ II
J I
fun revpoly0 ([], x) = 0.0
Page 411 of 709
| revpoly0 ((h::T), x) =
Go Back
h + x * revpoly0 (T, x)
Full Screen
Close
Quit
IIT Delhi
Tail Recursive S. Arun-Kumar, CSE
Title Page
revpoly1(L, x) = revpoly1 T R(L, x, 1, x)

Contents
where
JJ II
revpoly1 T R(L, x, p, s) =
J I
s if L = nil Page 412 of 709
revpoly1(T, x, px, s + ph) if L = h :: T Go Back
Full Screen
Close
Quit
Tail Recursion: SML IIT Delhi
local S. Arun-Kumar, CSE
fun revpoly1_TR([],x,p,s)=s Title Page
| revpoly1_TR((h::T),x,p,s)= Contents
revpoly1_TR(T,x,p*x,s+p*h) JJ II
in J I
fun revpoly1 (L, x) = Page 413 of 709
revpoly1_TR (L, x, 1.0, 0.0) Go Back
end Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Complexity of revpoly1 Title Page
Contents
Space. O(n) space to store the list JJ II
Multiplications. 2n multiplications J I
Additions. n additions. Page 414 of 709
Go Back
Full Screen
Close
Quit
Generating Primes
upto n IIT Delhi
Definition 12 A positive integer n > 1 S. Arun-Kumar, CSE
is composite iff it has a proper divisor Title Page
d|n with 1 < d < n. Otherwise it is Contents
prime. JJ II
• 2 is the smallest (first) prime. J I
Page 415 of 709
• 2 is the only even prime.

Go Back
• No other even number can be a Full Screen
prime. Close
• All other primes are odd Quit

More Properties
IIT Delhi
• An odd number cannot have any S. Arun-Kumar, CSE
even divisors.
Title Page
• Every number may be expressed Contents
uniquely (upto order) as a product JJ II
of prime factors. J I
• No divisor of a positive integer can Page 416 of 709
be greater than itself. Go Back
• For √ d|n such that d ≤

√ each divisor
Full Screen
b nc, n/d ≥ b nc is also a divisor. Close
Quit
Composites IIT Delhi
S. Arun-Kumar, CSE
• If a number n is composite, then it Title Page
√ a proper divisor d, 2 ≤ d ≤
has Contents
b nc. JJ II
• If a number n is composite, then√ it J I
has a prime divisor p, 2 ≤ p ≤ b nc. Page 417 of 709
• An odd composite number n has √ an

Go Back
odd prime divisor p, 3 ≤ p ≤ b nc. Full Screen
Close
Quit
Odd Primes IIT Delhi
• An odd number > 1 is a prime iff it S. Arun-Kumar, CSE
has no proper odd divisors Title Page
• An odd number > 1 is a prime iff

Contents
JJ II
it is not divisible by any odd prime
smaller than itself. J I
Page 418 of 709
• An odd number n > 1 is a prime iff

Go Back
it is√not divisible by any odd prime

Full Screen
≤ b nc.
Close
Quit
primesU pto(n) IIT Delhi
S. Arun-Kumar, CSE
primesU pto(n) =
Title Page


 [] if n < 2 Contents
 [(1, 2)] if n = 2

 JJ II
primesU pto(n − 1) elseif even(n) J I
generateF rom otherwise




 Page 419 of 709
([(1, 2)], 3, n, 2)

Go Back
where Full Screen
Close
Quit
generateF rom(P, m, n, k) IIT Delhi
bound function n − m S. Arun-Kumar, CSE
Invariant Title Page
2 < m ≤ n ∧ odd(m) Contents
JJ II
implies
J I
P = [(k − 1, pk−1), · · · , (1, p1)]
Page 420 of 709
and Go Back
∀q : pk−1 < q < m : composite(q) Full Screen
Close
Quit
generateF rom
IIT Delhi
generateF rom(P, m, n, k) =
S. Arun-Kumar, CSE


 P if m > n Title Page



 Contents
 generateF rom elseif



JJ II
(((k, m) :: P ), m + 2, n, k + 1) pwrt J I



 Page 421 of 709
generateF rom else




 Go Back
(P, m + 2, n, k)

Full Screen
where pwrt = primeW RT (m, P ) Close
Quit
primeW RT (m, P )
Definition 13 A number m is prime IIT Delhi
with respect to a list L of numbers iff it S. Arun-Kumar, CSE
is not divisible by any of them. Title Page
• A number is prime iff it is prime Contents
with respect to the list of all primes JJ II
smaller than itself. J I
• From properties of odd primes it fol-

Page 422 of 709
lows that a number n is prime iff it Go Back
is prime with
√ respect to the list of all
Full Screen
primes ≤ n Close
Quit
primeW RT (m, P ) IIT Delhi
bound function length(P ) S. Arun-Kumar, CSE
Invariant If P = [(i − 1, pi−1), . . . (1, p1)], Title Page
for some i ≥ 1 then Contents
• pk ≥ m > pk−1, and JJ II
J I
• m is prime with respect to
[(k − 1, pk−1), . . . , (i, pi)]
Page 423 of 709
Go Back
• m is a prime iff it is a prime with

Full Screen
respect to P
Close
Quit
primeW RT IIT Delhi
primeW RT (m, P ) = S. Arun-Kumar, CSE


 true if P = nil Title Page
f alse elseif h|m

 Contents

 primeW RT else JJ II
(m, tl(P ))
 J I
Page 424 of 709
where
Go Back
(i, h) = hd(P )
Full Screen
for some i > 0

Close
Quit
Density of Primes
Let π(n) denote the number of primes
IIT Delhi
upto n. Then S. Arun-Kumar, CSE
Title Page
n π(n) %
Contents
100 25 25.00%
JJ II
1000 168 16.80%
J I
10000 1229 12.29%
100000 9592 9.59% Page 425 of 709
1000000 78, 498 7.85% Go Back
10000000 664579 6.65% Full Screen
100000000 5761455 5.76% Close
Quit
The Prime Number
Theorem IIT Delhi
S. Arun-Kumar, CSE
π(n) Title Page
limn→∞ n/ ln n = 1
Contents
Proved by Gauss. JJ II
• Shows that the primes get sparser J I
at higher n Page 426 of 709
• A larger percentage of numbers as Go Back
we go higher are composite. Full Screen
from David Burton: Elementary Number Close
Theory. Quit
The Prime Number
Theorem
IIT Delhi
π(n)
n π(n) % lim S. Arun-Kumar, CSE
n→∞ n/ ln n Title Page
100 25 25.00% Contents
1000 168 16.80% 1.159 JJ II
10000 1229 12.29% 1.132 J I
100000 9592 9.59% 1.104 Page 427 of 709
1000000 78, 498 7.85% 1.084 Go Back
10000000 664579 6.65% 1.071 Full Screen
100000000 5761455 5.76% 1.061

Close
from David Burton: Elementary Number Quit
Theory.
IIT Delhi
S. Arun-Kumar, CSE
function calls Contents
primesU pto 1 JJ II
generateF rom n/2 J I

P n
primeW RT m=3,odd(m) π(m)
Page 428 of 709
Go Back
Full Screen
Close
Quit
Diagnosis
For each m ≤ n,
• P is in descending order of the IIT Delhi
primes S. Arun-Kumar, CSE
Title Page
• m is checked for divisibility π(m)

Contents
times
JJ II
• From properties of odd primes it J I
should not be necessary

√ to check Page 429 of 709
each m more than π(b mc) times Go Back
for divisibility.
Full Screen
• Organize P in ascending order in- Close
stead of descending. Quit
ascending-order
5.2. Compound Data & Lists
1. Compound Data
IIT Delhi
2. Recap: Tuples
3. Tuple: Formation S. Arun-Kumar, CSE
4. Tuples: Selection
Title Page
5. Tuples: Equality
6. Tuples: Equality errors Contents
7. Lists: Recap
JJ II
8. Lists: Append
9. cons vs. @
J I
10. Lists of Functions
11. Lists of Functions Page 430 of 709
12. Arithmetic Sequences

Go Back
13. Tail Recursion
14. Tail Recursion Invariant Full Screen
15. Tail Recursion

Close
16. Another Tail Recursion: AS3
17. Another Tail Recursion: AS3 iter Quit
18. AS3: Complexity
19. Generating Primes: 2
20. primes2U pto(n)
21. generate2F rom(P, m, n, k) IIT Delhi
22. generate2F rom

S. Arun-Kumar, CSE
23. prime2W RT (m, P )
24. prime2W RT Title Page
25. primes2: Complexity

Contents
26. primes2: Diagnosis
JJ II
J I
Page 431 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Compound Data Title Page
• Forming (compound) data struc- Contents
tures from simpler ones JJ II
J I
• Breaking up compound data into its
Page 432 of 709
components.
Go Back
Full Screen
Close
Quit
Recap: Tuples IIT Delhi
S. Arun-Kumar, CSE
formation : Cartesian products of Title Page
types Contents
selection : Selection of individual com- JJ II
ponents J I
equality : Equality checking Page 433 of 709
equality errors : Equality errors Go Back
Full Screen
forward to Lists
Close
Quit
Tuple: Formation IIT Delhi
Standard ML of New Jersey, S. Arun-Kumar, CSE
- val a = ("arun", 1<2, 2); Title Page
val a = ("arun",true,2) Contents
: string * bool * int JJ II
- val b = ("arun", true, 2); J I
val b = ("arun",true,2) Page 434 of 709
: string * bool * int Go Back
Full Screen
back
Close
Quit
Tuples: Selection
- #2 a;
val it = true : bool IIT Delhi
- #1 a; S. Arun-Kumar, CSE
val it = "arun" : string Title Page
- #3 a; Contents
- #4 a; J I
stdIn:1.1-1.5 Error: operator and ope

Page 435 of 709
operator domain: {4:’Y; ’Z} Go Back
operand: string * bool * in

Full Screen
in expression: Close
(fn {4=4,...} => 4) a Quit
back
Tuples: Equality IIT Delhi
- a = b; S. Arun-Kumar, CSE
val it = true : bool Title Page
- (1<2, true) = (1.0 < 2.0, true); Contents
val it = true : bool JJ II
- (true, 1.0 < 2.4) J I
= (1.0 < 2.4, true); Page 436 of 709
val it = true : bool Go Back
Full Screen
back Close
Quit
Tuples: Equality errors
- ("arun", (1, true)) IIT Delhi
= ("arun", 1, true); S. Arun-Kumar, CSE
stdIn:1.1-29.39 Error: operator andTitle Page

o
operator domain: (string * (int * Contents
b
operand: (string * (int * b
JJ II
in expression:
J I
("arun",(1,true))
Page 437 of 709
= ("arun",1,true)
Go Back
- ("arun", (1, true))
Full Screen
= (("arun", 1), true);
Close
stdIn:1.1-29.39 Error: operator and o
Quit
IIT Delhi
S. Arun-Kumar, CSE
Lists: Recap Title Page
formation : Sequence α List Contents
JJ II
selection : Selection of individual com-
J I
ponents
Page 438 of 709
new lists : Making new lists from old

Go Back
Full Screen
Close
Quit
Lists: Append
- op @;
IIT Delhi
val it = fn : ’a list * ’a list

S. Arun-Kumar, CSE
-> ’a list
Title Page
- [1,2,3] @ [˜1, ˜3];

Contents
val it = [1,2,3,˜1,˜3]
JJ II
: int list
J I
- [[1,2,3], [˜1, ˜2]]
@ [[1,2,3], [˜1, ˜2]]; Page 439 of 709
val it = Go Back
[[1,2,3], [˜1,˜2], Full Screen
[1,2,3], [˜1,˜2]] Close
: int list list Quit

cons vs. @ IIT Delhi
S. Arun-Kumar, CSE
cons is a constant time = O(1) oper-

Title Page
ation. But @ is linear = O(n) in the
length n of the first list. @ is defined
Contents
JJ II
as
L@M = J I
Page 440 of 709
M if L = nil
Go Back
h :: (T @M ) if L = h :: T
Full Screen
Close
Quit
IIT Delhi
Lists of Functions S. Arun-Kumar, CSE
- fun add1 x = x+1; Title Page
val add1 = fn : int -> int Contents
- fun add2 x = x + 2; JJ II
val add2 = fn : int -> int J I
- fun add3 x = x + 3; Page 441 of 709
val add3 = fn : int -> int Go Back
Full Screen
Close
Quit
Lists of Functions IIT Delhi
- val addthree S. Arun-Kumar, CSE
= [add1, add2, add3]; Title Page
val addthree Contents
= [fn,fn,fn] : (int -> int) list JJ II
- fun addall x = [(add1 x), (add2 x),

J I
val addall = fn : int -> int list

Page 442 of 709
- addall 3; Go Back
val it = [4,5,6] : int list Full Screen
Close
Quit
Arithmetic Sequences
AS1(a, d, n) = IIT Delhi

 [] if n ≤ 0 Title Page

Contents

 AS1(a, d, n − 1) else JJ II
@[a + (n − 1) ∗ d]

J I
function calls Order Page 443 of 709
AS1 n O(n) Go Back
@ n O(n) Full Screen
Pn 2)
:: i=0 i O(n Close
Quit
Tail Recursion IIT Delhi
AS2(a, d, n) = S. Arun-Kumar, CSE
 Title Page
 [] if n ≤ 0
Contents
JJ II
AS2 iter(a, d, n − 1, 0, []) else

J I
for any initial L0 and n ≥ k ≥ 0 Go Back
Full Screen
IN V 2 : L = L0@[a]@ . . . @[a + (k − 1) ∗ d]
Close
Quit
Tail Recursion:
Invariant IIT Delhi
S. Arun-Kumar, CSE
IN V 2 : L = L0@[a]@ . . . @[a + (k − 1) ∗ d]
Title Page
and bound function Contents
n−k JJ II
AS2 iter(a, d, n, k, L) = J I


 L if k ≥ n Page 445 of 709
 Go Back

 AS2 iter(a, d, n, k + 1 else Full Screen
L@[a + k ∗ d])
 Close
Quit
Tail Recursion: IIT Delhi
Complexity S. Arun-Kumar, CSE
Title Page
function calls Order Contents
AS2 1 JJ II
AS2 iter n O(n) J I
@ n O(n) Page 446 of 709

Pn 2)
:: i=0 i O(n Go Back
So tail recursion simply doesn’t help! Full Screen
Close
Quit
Another Tail Recursion
: AS3 IIT Delhi
AS3(a, d, n) = S. Arun-Kumar, CSE

 [] if n ≤ 0 Title Page
Contents
AS3 iter(a, d, n − 1, []) else

 JJ II
J I
where
Page 447 of 709
for any initial L0, n0 ≥ n > 0, and

Go Back
IN V 3 : L = (a + (n − 1) ∗ d) :: · · · :: Full Screen
Close
· · · :: (a + (n0 − 1) ∗ d) :: L0
Quit
Another Tail Recursion:
AS3 iter IIT Delhi
IN V 3 : L = (a + (n − 1) ∗ d) :: · · · S. Arun-Kumar, CSE
· · · :: (a + (n0 − 1) ∗ d) :: L0 Title Page
and bound function n, Contents
AS3 iter(a, d, n, L) = JJ II
 J I

 L if n ≤ 0 Page 448 of 709

Go Back

 AS3 iter(a, d, n − 1, else
Full Screen
(a + (n − 1) ∗ d)::L)

Close
Quit
IIT Delhi
AS3: Complexity S. Arun-Kumar, CSE
Title Page
function calls Order Contents
AS3 1 JJ II
AS3 iter n O(n) J I
@ 0 Page 449 of 709

P n
:: i=0 1 O(n) Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Generating Primes: 2 Title Page
composites Contents
• primesU pto JJ II
J I
• invariant
Page 450 of 709
• generateF rom Go Back
Full Screen
Close
Quit
primes2U pto(n) IIT Delhi
S. Arun-Kumar, CSE
primes2U pto(n) =
Title Page


 [] if n < 2 Contents
 [(1, 2)] if n = 2

 JJ II
primes2U pto(n − 1) elseif even(n) J I
generate2F rom otherwise




 Page 451 of 709
([(1, 2)], 3, n, 2)

Go Back
where Full Screen
Close
Quit
generate2F rom(P, m, n, k) IIT Delhi
bound function n − m S. Arun-Kumar, CSE
Invariant Title Page
2 < m ≤ n ∧ odd(m) Contents
JJ II
implies
J I
P = [(1, p1), · · · , (k − 1, pk−1)]
Page 452 of 709
and Go Back
∀q : pk−1 < q < m : composite(q) Full Screen
Close
Quit
generate2F rom
IIT Delhi
generate2F rom(P, m, n, k) =
S. Arun-Kumar, CSE


 P if m > n Title Page



 Contents
 generate2F rom elseif



JJ II
(P @[(k, m)], m + 2, n, k + 1) pwrt J I




 Page 453 of 709
generate2F rom else




 Go Back
(P, m + 2, n, k)

Full Screen
where pwrt = prime2W RT (m, P ) Close
Quit
prime2W RT (m, P )
IIT Delhi
bound function length(P )
S. Arun-Kumar, CSE
Invariant If P = [(i, pi), . . . (k − 1, pk−1)], Title Page
for some i ≥ 1 then Contents
• pk ≥ m > pk−1, and JJ II
• m is prime with respect to J I
[(1, p1), . . . , (i − 1, pi−1)] Page 454 of 709
• m is a prime iff it is a prime Go Back
with respect √ to [(1, p1), . . . , (j, pj )], Full Screen
where pj ≤ b mc < pj+1 Close
Quit
prime2W RT
IIT Delhi
prime2W RT (m, P ) =
S. Arun-Kumar, CSE


 true if P = nil Title Page
 true if h > m div h


 Contents
f alse elseif h|m JJ II
primeW RT else



 J I
(m, tl(P ))

Page 455 of 709
where Go Back
(i, h) = hd(P ) Full Screen
for some i > 0 Close
Quit
IIT Delhi
: Complexity
S. Arun-Kumar, CSE
primes2 Title Page
Contents
function calls JJ II
primes2U pto 1 J I
generate2F rom n/2 √
P n Page 456 of 709
prime2W RT m=3,odd(m) π(b mc) Go Back
Full Screen
Close
Quit
primes2 : Diagnosis
generate2F rom IIT Delhi
• Uses @ to create an ascending se- S. Arun-Kumar, CSE
quence of primes Title Page
• For each new prime pk this opera- Contents
tion takes time O(k). JJ II
J I
• Can tail recursion be used to re-
Page 457 of 709
duce the complexity due to @?

Go Back
• Can a more efficient algorithm us- Full Screen
ing :: instead of @ be devised (as in Close
the case of AS3)? Quit

5.3. Compound Data & List Algorithms
1. Compound Data: Summary
IIT Delhi
2. Records: Constructors
3. Records: Example 1 S. Arun-Kumar, CSE
4. Records: Example 2
Title Page
5. Records: Destructors
6. Records: Equality Contents
7. Tuples & Records

JJ II
8. Back to Lists
9. Lists: Correctness
J I
10. Lists: Case Analysis
11. Lists: Correctness by Cases Page 458 of 709
12. List-functions: length

Go Back
13. List Functions: search
14. List Functions: search2 Full Screen
15. List Functions: ordered

Close
16. List Functions:insert
17. List Functions: reverse Quit
18. List Functions: reverse2
19. List Functions:merge
21. List Functions:merge contd. IIT Delhi
22. ML: merge

S. Arun-Kumar, CSE
23. Sorting by Insertion
24. Sorting by Merging Title Page
25. Sorting by Merging

Contents
26. Functions as Data
27. Higher Order Functions JJ II
J I
Page 459 of 709
Go Back
Full Screen
Close
Quit
Compound Data:
Summary IIT Delhi
• Compound Data: S. Arun-Kumar, CSE
Title Page
Tuples: Cartesian products of differ-

Contents
ent types (ordered)

JJ II
Lists: Sequences of the same type J I
of element
Page 460 of 709
Records: Unordered named aggre- Go Back
gations of elements of different Full Screen
types.
Close
• Constructors & Destructors Quit

Records: Constructors
• A record is a set of values drawn IIT Delhi
from various types such that each S. Arun-Kumar, CSE
component (called a field) has a Title Page
unique name. Contents
• Each record has a type defined by JJ II
J I
field names
Page 461 of 709
types of fieldnames
Go Back
The order of presentation of the Full Screen
record fields does not affect its Close
type in any way.

Quit
Records: Example 1
- val pinky = S. Arun-Kumar, CSE
{ name = "Pinky", age = 3, Title Page
fav_colour = "pink"}; Contents
- val pinky = {age=3,

JJ II
fav_colour="pink",
J I
name="Pinky"}
Page 462 of 709
: {age:int,
Go Back
fav_colour:string,
Full Screen
name:string
Close
}
Quit
Records: Example 2
- val billu = IIT Delhi
{ age = 1, S. Arun-Kumar, CSE
name = "Billu", Title Page
fav_colour = "blue" Contents
}; JJ II
- val billu = J I
{age=1,fav_colour="blue",name="Billu"
Page 463 of 709
{age:int, fav_colour:string, name:str Go Back
- pinky = billu; Full Screen
val it = false : bool Close
Quit
IIT Delhi
Records: Destructors S. Arun-Kumar, CSE
#age billu; Title Page
val it = 1 : int Contents
- #fav_colour billu; JJ II
val it = "blue" : string J I
- #name billu; Page 464 of 709
val it = "Billu" : string Go Back
Full Screen
Close
Quit
Records: Equality
- val pinky2 = IIT Delhi
{ name = "Pinky", S. Arun-Kumar, CSE
fav_colour = "pink", Title Page
age = 3 Contents
}; JJ II
- val pinky2 = J I
{age=3,fav_colour="pink",name="Pinky"
Page 465 of 709
{age:int, fav_colour:string, name:str Go Back
- pinky = pinky2; Full Screen
val it = true : bool Close
Quit
Tuples & Records IIT Delhi
S. Arun-Kumar, CSE
• A k-tuple may be thought of as a

Title Page
record whose fields are numbered
#1 to #k instead of having names.
Contents
JJ II
• A record may be thought of as J I
a generalization of tuples whose
Page 466 of 709
components are named rather than

Go Back
being numbered.
Full Screen
back
Close
Quit
Back to Lists
IIT Delhi
• Every L : α List satisfies

S. Arun-Kumar, CSE
L = []
Title Page
XOR Contents
L = hd(L) :: tl(L) JJ II
• Many functions on lists (L) are de- J I
fined by induction on its length (|L|). Page 467 of 709
Go Back
c if L = []
f (L) = Full Screen
g(h, T ) if L = h :: T Close
Quit
Lists: Correctness IIT Delhi
S. Arun-Kumar, CSE
Hence their properties (P ) are proved Title Page
by induction on the length of the list. Contents
Basis |L| = 0. Prove P ([]) JJ II
Induction hypothesis (IH) Assume for J I
some |T | = n > 0, P (T ) holds. Page 468 of 709
Induction Step Prove P (h :: T ) for L = Go Back
h :: T with |L| = n + 1 Full Screen
Close
Quit
Lists: Case Analysis
inductive defns on lists IIT Delhi
• Every list has exactly one of the fol- S. Arun-Kumar, CSE
lowing forms (patterns) Title Page
– [] Contents
JJ II
– h::T
J I
• ML provides convenient
Page 469 of 709
case analysis based on patterns.

Go Back
fun f [] = c Full Screen
| f (h::T) = g (h, T) Close
; Quit
Lists: Correctness by
Cases IIT Delhi
Lists-correctness S. Arun-Kumar, CSE
Title Page
P is proved by case analysis.

Contents
Basis Prove JJ II
P ([])
J I
Induction hypothesis (IH) Assume Page 470 of 709
P (T ) Go Back
Induction Step Prove Full Screen
P (h :: T ) Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
List-functions: length
Title Page
Contents

length [] = 0 JJ II
length (h :: T ) = 1 + (length T ) J I
Page 471 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
List Functions: search S. Arun-Kumar, CSE
Title Page
To determine whether x occurs in a list

Contents
L
 JJ II
 search (x, []) = f alse J I
search (x, h :: T ) = true if x = h Page 472 of 709
search (x, h :: T ) = search(x, T ) else


Go Back
Full Screen
Close
Quit
IIT Delhi
List Functions: search2 S. Arun-Kumar, CSE
Title Page
Or even more conveniently

Contents

 search2 (x, []) = f alse JJ II
search2 (x, h :: T ) = (x = h) or J I
search2 (x, T )

Page 473 of 709
Time Complexity?? Go Back
Full Screen
Close
Quit
List Functions: ordered
IIT Delhi
Definition 14 A list L = [a0, . . . , an−1]

S. Arun-Kumar, CSE
is ordered by a relation ≤ if consec-
utive elements are related by ≤, i.e
Title Page
ai ≤ ai+1, for 0 ≤ i < n − 1. Contents
JJ II


 ordered [] J I
ordered [h]

Page 474 of 709

 ordered (h0 :: h1 :: T ) if h0 ≤ h1 and Go Back
ordered(h1 :: T )

Full Screen
Time Complexity?? Close
Quit
List Functions:insert
Given an ordered list L : α List, insert IIT Delhi
an element x : α at an appropriate po- S. Arun-Kumar, CSE
sition Title Page



 insert (x, []) = [x] Contents



 JJ II
 insert (x, h :: T ) = x :: (h :: T )


J I
if x ≤ h Page 475 of 709




 Go Back
insert (x, h :: T ) = h :: (insert (x, T ))




 Full Screen
else

Close
Time Complexity?? Quit

List Functions:
IIT Delhi
reverse S. Arun-Kumar, CSE
Reverse the elements of a list Title Page
L = [a0, . . . , an−1] to obtain M = Contents
[an−1, . . . , a0]. JJ II

reverse [] = [] J I
reverse (h :: T ) = (reverse T )@[h] Page 476 of 709
Go Back
Time Complexity?? O(n2) Full Screen
Close
Quit
List Functions: reverse2 IIT Delhi
S. Arun-Kumar, CSE

reverse [] = [] Title Page
reverse (h :: T ) = rev ((h :: T ), []) Contents
where JJ II

rev ([], N ) = N J I
rev (h :: T, N ) = rev (T, h :: N ) Page 477 of 709
Go Back
Correctness ??
Time Complexity?? O(n) Full Screen
Close
Quit
List Functions:merge IIT Delhi
S. Arun-Kumar, CSE
Merge two ordered lists |L| = l, |M | =

Title Page
m to produce an ordered list |N | = l +

Contents
m containing exactly the elements of
JJ II
L and M . That is if
L = [1, 3, 5, 9, 11] and J I
M = [0, 3, 4, 4, 10], then Page 478 of 709
merge(L, M ) = N , where Go Back
N = [0, 1, 3, 3, 4, 4, 5, 9, 10, 11] Full Screen
Close
Quit
List Functions:merge
IIT Delhi


 merge([], M ) = M S. Arun-Kumar, CSE


 Title Page

merge(L, []) = L

 Contents



 JJ II

merge(L, M ) = J I
a :: (merge(S, M )) if a ≤ b



 Page 479 of 709



 Go Back
merge(L, M ) =



 Full Screen
b :: (merge(L, T )) else

Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
List Functions:merge Title Page
contd. Contents
JJ II
where
L = a :: S J I
M = b :: T Page 480 of 709
Go Back
Full Screen
Close
Quit
ML: merge IIT Delhi
S. Arun-Kumar, CSE
fun merge ([], M) = M Title Page
| merge (L, []) = L

Contents
| merge (L as a::S,
JJ II
M as b::T) =
J I
if a <= b
Page 481 of 709
then a::merge(S, M)
Go Back
else b::merge(L, T)
Full Screen
Close
Quit
Sorting by Insertion
Given a list of elements to reorder IIT Delhi
them (i.e. with the same number of S. Arun-Kumar, CSE
occurrences of each element as in the Title Page
original list) to produce a new ordered Contents
list. JJ II
Hence sort[10, 8, 3, 6, 9, 7, 4, 8, 1] = J I
[1, 3, 4, 6, 7, 8, 8, 9, 10] Page 482 of 709

isort[] = [] Go Back
isort(h :: T ) = insert(h, (isortT )) Full Screen
Close
Time Complexity??
Quit
Sorting by Merging IIT Delhi
S. Arun-Kumar, CSE


 msort [] = [] Title Page
msort [a] = [a]


Contents

 msort L = merge ((msort M ), JJ II
(msort N ))

J I
Go Back
(M, N ) = split L
Full Screen
Close
Quit
Sorting by Merging IIT Delhi
where S. Arun-Kumar, CSE

 split [] = ([], []) Title Page
split [a] = ([a], []) Contents
split (a :: b :: P ) = (a :: Lef t, b :: Right)

 JJ II
J I
(Lef t, Right) = split P Go Back
Full Screen
Time Complexity?? Close
Quit
Functions as Data IIT Delhi
list of functions
S. Arun-Kumar, CSE
• Every function is unary. A function Title Page
of many arguments may be thought Contents
of as a function of a single argu- JJ II
ment i.e. a tuple of appropriate J I
type. Page 485 of 709
• Every function is a value of an ap- Go Back
propriate type. Full Screen
• Hence functions are also data. Close
Quit
Higher Order IIT Delhi
Functions S. Arun-Kumar, CSE
Compound data may be constructed Title Page
from functions as values using the Contents
constructors of the compound data JJ II
structure. J I
Page 486 of 709
Functions may be defined with

Go Back
other functions and/or data as arguments
Full Screen
to produce new values or new functions.
Close
Quit
6. Higher Order Functions & Structured Data
6.1. Higher Order Functions
1. Summary: Compound Data
IIT Delhi
2. List: Examples
3. Lists: Sorting S. Arun-Kumar, CSE
4. Higher Order Functions

Title Page
5. An Example
6. Currying
Contents
7. Currying: Contd
8. Generalization JJ II
9. Generalization: 2
J I
10. Applying a list
11. Trying it out Page 487 of 709
12. Associativity
Go Back
13. Apply to a list
14. Sequences Full Screen
15. Further Generalization

Close
17. Sequences
Quit
18. Efficient Generalization
19. Sequences: 2
20. More Generalizations
21. More Summations IIT Delhi
22. Or Maybe . . . Products

N S. Arun-Kumar, CSE
23. Or Some Other
N
24. Other Title Page
25. Examples of ⊗, e
Contents
JJ II
J I
Page 488 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
Summary: Compound S. Arun-Kumar, CSE
Data Title Page
Contents
• Records and tuples JJ II
• Lists J I
– Correctness Page 489 of 709
– Examples Go Back
Full Screen
Close
Quit
IIT Delhi
List: Examples S. Arun-Kumar, CSE
Title Page
• Length of a list
Contents
• Searching a list JJ II
• Checking whether a list is ordered J I
• Reversing a list Page 490 of 709
• Sorting of lists Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Title Page
Lists: Sorting Contents
• Sorting by insertion JJ II
J I
• Sorting by Divide-and-Conquer
Page 491 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Higher Order Title Page
Functions JJ
Contents
II
• Functions as data J I
• Higher order functions Page 492 of 709
Go Back
Full Screen
Close
Quit
An Example IIT Delhi
List of functions S. Arun-Kumar, CSE
Title Page
• add1 x = x + 1
Contents
• add2 x = x + 2 JJ II
• add3 x = x + 3 J I
Suppose we needed to define a long Page 493 of 709
list of length n, where the i-th element Go Back
is the function that adds i + 1 to the

argument. Full Screen
Close
Quit
Currying
addc y x = x + y IIT Delhi
S. Arun-Kumar, CSE
ML’s response : Title Page
val addc = fn : Contents
int -> (int -> int)

JJ II
Contrast with ML’s response J I
- op +; Page 494 of 709
val it = fn : int * int -> int Go Back
addc is the curried version of the bi- Full Screen
nary operation +. Close
Quit
Currying: Contd
IIT Delhi
S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 495 of 709
Go Back
Full Screen
Close
Quit
Generalization IIT Delhi
Then S. Arun-Kumar, CSE
Title Page
• addc1 = (addc 1): int -> int

Contents
• addc2 = (addc 2): int -> int JJ II
• addc3 = (addc 3): int -> int J I
and for any i, Page 496 of 709
(addc i): int -> int Go Back
Full Screen
is the required function.
Close
Quit
IIT Delhi
Generalization: 2 S. Arun-Kumar, CSE
list adds n = Title Page

[] if n ≤ 0 Contents
(list adds(n − 1))@[(addc n)] else JJ II
J I
ML’s response : Page 497 of 709
val list_adds = fn : Go Back
int -> (int -> int) list Full Screen
Close
Quit
Applying a list IIT Delhi
addall S. Arun-Kumar, CSE
Title Page
applyl [] x = [] Contents
applyl (h :: T ) x = (h x) :: (applyl T x) JJ II
J I
ML’s response:
Page 498 of 709
val applyl = fn :
Go Back
(’a -> ’b) list ->

Full Screen
’a -> ’b list
Close
Quit
IIT Delhi
Trying it out S. Arun-Kumar, CSE
interval x n = applyl x (list adds n) Title Page
ML’s response: Contents
val interval = fn : JJ II
int -> int -> int list J I
- interval 53 5; Page 499 of 709
val it = [54,55,56,57,58] Go Back
: int list Full Screen
Close
Quit
Associativity IIT Delhi
• Application associates to the left. S. Arun-Kumar, CSE
f x y = ((f x) y) Title Page
• → associates to the right.

Contents
α → β → γ = α → (β → γ) JJ II
J I
If f : α → β → γ → δ Page 500 of 709
then f a : β → γ → δ Go Back
and f a b : γ → δ Full Screen
and f a b c : δ Close
Quit
Apply to a list
Apply a list Transpose of a matrix IIT Delhi

map f [] = [] S. Arun-Kumar, CSE
map f (h :: T ) = (f h) :: (map f T ) Title Page
Contents
val it = fn : (’a -> ’b) -> JJ II
’a list -> ’b list J I
- map addc3 [4, 6, ˜1, 0]; Page 501 of 709
val it = [7,9,2,3] : int list Go Back
- map real [7,9,2,3]; Full Screen
val it = [7.0,9.0,2.0,3.0] Close
: real list Quit

Sequences IIT Delhi
Arithmetic sequences-1 S. Arun-Kumar, CSE
Arithmetic sequences-2
Title Page
Arithmetic sequences-3
Contents
AS4(a, d, n) =
JJ II


 [] if n ≤ 0 J I

Page 502 of 709

 a :: (map (addc d) else Go Back
(AS4 (a, d, (n − 1))))


Full Screen
Close
Quit
Further Generalization
Given IIT Delhi
f :α∗α→α S. Arun-Kumar, CSE
Then Title Page
curry2 f x y = f (x, y) Contents
JJ II
and J I
(curry2 f ) : α → (α → α) Page 503 of 709
Go Back
and for any d : α,

Full Screen
((curry2 f ) d) : α → α Close
Quit
Further Generalization IIT Delhi
S. Arun-Kumar, CSE
seq(f, a, d, n) =
Title Page


 [] if n ≤ 0 Contents

JJ II

 a :: (map ((curry2 f ) d) J I
(seq (f, a, d, n − 1))) else


Page 504 of 709
is the sequence of length n generated Go Back
with ((curry2 f ) d), starting from a. Full Screen
Close
Quit
Sequences IIT Delhi
Arithmetic: AS5(a, d, n) = S. Arun-Kumar, CSE
seq(op+, a, d, n) Title Page
Geometric: GS1(a, r, n) = Contents
seq(op∗, a, r, n) JJ II
Harmonic: HS1(a, d, n) = J I
map reci (AS5(a, d, n)) Page 505 of 709
Go Back
where
reci x = 1.0/(real x) gives the recipro- Full Screen
cal of a (non-zero) integer. Close
Quit
Efficient Generalization IIT Delhi
Let’s not use map repeatedly. S. Arun-Kumar, CSE
seq2(f, g, a, d, n) = Title Page

 [] if n ≤ 0
Contents
JJ II
(f a) :: (seq2 (f, g(a, d), d, n − 1)) else J I


Page 506 of 709
is the sequence of length n generated Go Back
with a unary f , a binary g starting from Full Screen
f (a). Close
Quit
IIT Delhi
Sequences: 2
S. Arun-Kumar, CSE
Title Page
• AS6(a, d, n) = seq2(id, op+, a, d, n) Contents
• GS2(a, r, n) = seq2(id, op∗, a, r, n) JJ II
J I
• HS2(a, d, n) = seq2(reci, op+, a, d, n)
Page 507 of 709
where id x = x is the identity function. Go Back
Full Screen
Close
Quit
More Generalizations
Often interested in some particular IIT Delhi
measure related to a sequence, rather S. Arun-Kumar, CSE
than in the sequence itself, e.g. sum- Title Page
mations of Contents
• arithmetic, geometric, harmonic se- JJ II
quences J I
• ex,
Page 508 of 709
trigonometric functions upto

Go Back
some n-th term

Full Screen
• (Truncated) Taylor and Maclaurin Close
series Quit
More Summations
Wasteful to first generate the se-
quence and then compute the mea- IIT Delhi
sure S. Arun-Kumar, CSE
Xu Title Page
f (i) Contents
i=l JJ II
J I
where the range [l, u] is defined by a
unary succ function Page 509 of 709
sum(f, succ, l, u) = Go Back
Full Screen
0 if [l, u] = ∅ Close
f (l)+sum(f, succ, succ(l), u) else

Quit
Or Maybe . . . Products IIT Delhi
Or may be interested in forming prod- S. Arun-Kumar, CSE
ucts of sequences. Title Page
u
Y Contents
f (i) JJ II
i=l J I
prod(f, succ, l, u) = Page 510 of 709

1 if [l, u] = ∅ Go Back
f (l)∗prod(f, succ, succ(l), u) else Full Screen
Close
Quit
Or Some Other N
Or some other binary operation ⊗

IIT Delhi
which has the following properties:

S. Arun-Kumar, CSE
• ⊗ : (α ∗ α) → α is closed Title Page
• ⊗ is associative i.e. Contents
a ⊗ (b ⊗ c) = (a ⊗ b) ⊗ c JJ II
• ⊗ has an identity element e i.e J I
a⊗e=a=e⊗a Page 511 of 709
Go Back
u
O Full Screen
f (l) Close
i=l Quit
IIT Delhi
Other N S. Arun-Kumar, CSE
Title Page
Then if f, succ : α → α Contents
ser(⊗, f, succ, l, u) = JJ II

e if [l, u] = ∅ J I
f (l) ⊗ ser(⊗, f, succ, succ(l), u) else Page 512 of 709
Go Back
Full Screen
Close
Quit
Examples of ⊗, e IIT Delhi
• +, 0 on integers and reals S. Arun-Kumar, CSE
Title Page
• concatenation and the empty string Contents
on strings
JJ II
• andalso, true on booleans J I
• orelse, false on booleans Page 513 of 709
• +, 0 on vectors and matrices Go Back
• ∗, 1 on vectors and matrices

Full Screen
Close
Quit
6.2. Structured Data
1. Transpose of a Matrix
2. Transpose: 0
3. Transpose: 10 IIT Delhi
4. Transpose: 01
S. Arun-Kumar, CSE
5. Transpose: 20
6. Transpose: 02 Title Page
7. Transpose: 30
Contents
8. Transpose: 03
9. trans JJ II
10. is2DM atrix
11. User Defined Types J I
12. Enumeration Types

Page 514 of 709
13. User Defined Structural Types
14. Functions vs. data Go Back
15. Data as 0-ary Functions

Full Screen
16. Data vs. Functions
17. Data vs. Functions: Recursion Close
18. Lists
Quit
19. Constructors
20. Shapes
21. Shapes: Triangle Inequality
22. Shapes: Area IIT Delhi
23. Shapes: Area

S. Arun-Kumar, CSE
24. ML: Try out
25. ML: Try out (contd.) Title Page

Contents
27. Recursive Data Types
28. Resistors: Datatype JJ II
29. Resistors: Equivalent
30. Resistors J I
31. Resistors: Example

Page 515 of 709
32. Resistors: ML session
Go Back
Full Screen
Close
Quit
Transpose of a Matrix IIT Delhi
Map S. Arun-Kumar, CSE
Assume a 2-D r × c matrix is repre- Title Page
sented by a list of lists of elements. Contents
Then JJ II
transpose L = J I

trans L if is2DM atrix(L) Page 516 of 709
⊥ else Go Back
Full Screen
where Close
Quit
Transpose: 0
IIT Delhi
[ [ S. Arun-Kumar, CSE
[ 11 12 13 ]
Title Page
Contents
[ 21 22 23 ] JJ II
[ 31 32 33 ] J I
[ 41 42 43 ] Page 517 of 709
Go Back
] ]
Full Screen
Close
Quit
Transpose: 10
IIT Delhi
[ 11 12 13 ]
Title Page
Contents
[ 21 22 23 ] JJ II
[ 31 32 33 ] J I
[ 41 42 43 ] Page 518 of 709
Go Back
] ]
Full Screen
Close
Quit
Transpose: 01
IIT Delhi
[ 12 13 ] [ 11 21 31 41 ]
Title Page
Contents
[ 22 23 ] JJ II
[ 32 33 ] J I
[ 42 43 ] Page 519 of 709
Go Back
] ]
Full Screen
Close
Quit
Transpose: 20
IIT Delhi
[ 12 13 ] [ 11 21 31 41 ]
Title Page
Contents
[ 22 23 ] JJ II
[ 32 33 ] J I
[ 42 43 ] Page 520 of 709
Go Back
] ]
Full Screen
Close
Quit
Transpose: 02
IIT Delhi
[ 13 ] [ 11 21 31 41 ]
Title Page
Contents
[ 23 ]
[ 12 22 32 42 ] JJ II
[ 33 ] J I
[ 43 ] Page 521 of 709
Go Back
] ]
Full Screen
Close
Quit
Transpose: 30
IIT Delhi
[ 13 ] [ 11 21 31 41 ]
Title Page
Contents
[ 23 ]
[ 12 22 32 42 ] JJ II
[ 33 ] J I
[ 43 ] Page 522 of 709
Go Back
] ]
Full Screen
Close
Quit
Transpose: 03
IIT Delhi
[ ] [ 11 21 31 41 ]
Title Page
Contents
[ ]
[ 12 22 32 42 ] JJ II
[ ] J I
[ ] [ 13 23 33 43 ] Page 523 of 709
Go Back
] ]
Full Screen
Close
Quit
trans IIT Delhi

 trans [] = [] Title Page
trans [] :: T L = []

Contents

 trans LL = (map hd LL) :: JJ II
(trans (map tl LL))


J I
and Page 524 of 709
is2DM atrix = #1(dimensions L) Go Back
where Full Screen
Close
Quit
is2DM atrix
IIT Delhi


 dimensions [] = (true, 0, 0) S. Arun-Kumar, CSE



 Title Page
 dimensions [H] =


Contents
(true, 1, h) JJ II



 J I
dimensions (H :: T L) =



 Page 525 of 709
(b and (h = c), r + 1, c)

Go Back
where dimensions T L = (b, r, c) Full Screen
and h = length H Close
Quit
User Defined Types IIT Delhi
S. Arun-Kumar, CSE
Many languages allow user-defined Title Page
data types.
Contents
• record types: Pinky and Billu JJ II
• Enumerations: aggregates of het- J I
erogeneous data. Page 526 of 709
• other structural constructions (if Go Back
desperate!) Full Screen
Close
Quit
Enumeration Types
Many languages allow user-defined IIT Delhi
data types. S. Arun-Kumar, CSE
• record types: Pinky and Billu Title Page
• Enumerations: aggregates of het- Contents
erogeneous data. JJ II
J I
– days of the week
Page 527 of 709
– colours
Go Back
– geometrical shapes Full Screen
• other structural constructions (if Close
desperate!) Quit
User Defined
Structural Types
IIT Delhi
Many languages allow user-defined
data types. S. Arun-Kumar, CSE
Title Page
• record types: Pinky and Billu

Contents
• Enumerations: aggregates of het- JJ II
erogeneous data. J I
• other structural constructions (if Page 528 of 709
desperate!) Go Back
– trees Full Screen
– graphs Close
– symbolic expressions Quit

IIT Delhi
Functions vs. data S. Arun-Kumar, CSE
• Inspired by the list constructors, nil Title Page
and cons Contents
JJ II
• Grand Unification of functions and
J I
data
Page 529 of 709
– Functions as data
Go Back
– Data as functions Full Screen
Close
Quit
Data as 0-ary Functions
IIT Delhi
• Every data element may be re-

S. Arun-Kumar, CSE
garded as a function with 0 argu-

Title Page
ments
Contents
– Caution: A constant function JJ II

f (x) = 5, for all x : α
J I
where
f : α → int Page 530 of 709
is not the same as a value Go Back
5 : int Full Screen
. Their types are different. Close
Quit
IIT Delhi
Data vs. Functions S. Arun-Kumar, CSE
Title Page
Facilities Functions Data Contents
primitive operations values JJ II
user-defined functions constructors J I
composition application alternative Page 531 of 709
recursion recursion recursion Go Back
Full Screen
Close
Quit
IIT Delhi
Data vs. Functions: S. Arun-Kumar, CSE
Recursion Title Page
Recursion Contents
Basis JJ II
naming J I
composition Page 532 of 709
induction Go Back
Full Screen
Close
Quit
Lists as Structured
Data
datatype ’a list = IIT Delhi
nil | S. Arun-Kumar, CSE
cons of ’a * ’a list Title Page
Contents
Every α list is either JJ II
J I
nil : (Basis, name)
Page 533 of 709
| : or (alternative) Go Back
cons : constructed inductively from Full Screen
an element of type ’a and another Close
list of type ’a list using the con- Quit
structor cons
Constructors
IIT Delhi
• Inspired by the list constructors

S. Arun-Kumar, CSE
nil : α list Title Page
cons : α × α list → α list Contents
• combine heterogeneous types: α JJ II
and α list J I
• allows recursive definition by a form Page 534 of 709
of induction Go Back
Basis : nil Full Screen
Close
Induction : cons
Quit
Shapes IIT Delhi
S. Arun-Kumar, CSE
A non-recursive data type

Title Page
datatype shape =
Contents
JJ II
CIRCLE of real
J I
| RECTANGLE of real * real
Page 535 of 709
| TRIANGLE of
Go Back
real * real * real
Full Screen
Close
Quit
Shapes: Triangle IIT Delhi
Inequality S. Arun-Kumar, CSE
Title Page
fun isTriangle Contents
(TRIANGLE (a, b, c)) = JJ II
(a+b>c) andalso J I
(b+c>a) andalso Page 536 of 709
(c+a>b) Go Back
| isTriangle _ = false
Full Screen
Close
Quit
Shapes: Area IIT Delhi
exception notShape; S. Arun-Kumar, CSE
Title Page
fun area (CIRCLE (r)) = Contents
3.14159 * r * r JJ II
| area (RECTANGLE (l,b)) = J I
l*b Page 537 of 709
| area (s as Go Back
TRIANGLE (a, b, c)) = Full Screen
Close
Quit
Shapes: Area
IIT Delhi
S. Arun-Kumar, CSE
if isTriangle (s) then Title Page
let val s = (a+b+c)/2.0 Contents
in Math.sqrt JJ II
(s*(s-a)*(s-b)*(s-c)) J I
end Page 538 of 709
else raise notShape; Go Back
Full Screen
Close
Quit
ML: Try out
- use "shapes.sml"; IIT Delhi
[opening shapes.sml] S. Arun-Kumar, CSE
datatype shape Title Page
= CIRCLE of real Contents
| RECTANGLE of real * real

JJ II
| TRIANGLE of
J I
real * real * real
Page 539 of 709
val isTriangle =
Go Back
fn : shape -> bool
Full Screen
exception notShape
Close
val area = fn : shape -> real
Quit
ML: Try out (contd.)
val it = () : unit IIT Delhi
- area (TRIANGLE (2.0,1.0,3.0)); S. Arun-Kumar, CSE
Title Page
uncaught exception notShape Contents
raised at: shapes.sml:22.17-22.25 JJ II
- area J I
(TRIANGLE (3.0, 4.0, 5.0)); Page 540 of 709
- Full Screen
Close
Back to User defined types

Quit
Enumeration Types
• Enumeration types are non- IIT Delhi
recursive datatypes with S. Arun-Kumar, CSE
Title Page
• 0-ary constructors
Contents
datatype working = MON | TUE JJ II
| WED | THU | FRI; J I
datatype weekends = SAT | SUN Page 541 of 709
datatype weekdays = working Go Back
| weekends; Full Screen
Close
Back to User defined types

Quit
Recursive Data Types IIT Delhi
• But the really interesting types are S. Arun-Kumar, CSE
the recursive data types Title Page
Back to Lists Contents
JJ II
• As with lists proofs of correctness J I

on recursive data types depend on Page 542 of 709
a case-analysis of the structure (ba-

Go Back
sis and inductive constructors)

Full Screen
Correctness on lists Close
Quit
IIT Delhi
Resistors: Datatype S. Arun-Kumar, CSE
Title Page
datatype resist = Contents
RES of real |
JJ II
SER of resist * resist |
J I
PAR of resist * resist
Page 543 of 709
Go Back
Full Screen
Close
Quit
Resistors: Equivalent
IIT Delhi
fun value (RES (r)) = r S. Arun-Kumar, CSE
| value (SER (R1, R2)) = Title Page
value (R1) + value (R2) Contents
| value (PAR (R1, R2)) =

JJ II
let val r1 = value (R1);
J I
val r2 = value (R2)
Page 544 of 709
in (r1*r2)/(r1+r2)
Go Back
end;
Full Screen
Close
Quit
Resistors
5.0 IIT Delhi
5.0 2.0 S. Arun-Kumar, CSE
Title Page
4.0 Contents
JJ II
J I
Page 545 of 709
3.0 Go Back
Full Screen
+ − Close
Quit
Resistors: Example
val R = PAR(
SER(
IIT Delhi
PAR(
S. Arun-Kumar, CSE
RES (5.0),
Title Page
RES (4.0)
Contents
),
JJ II
SER(
J I
RES (5.0),
RES (2.0) Page 546 of 709
) Go Back
), Full Screen
RES(3.0) Close
); Quit
Resistors: ML session
- use "resistors.sml";
[opening resistors.sml] IIT Delhi
datatype resist = PAR of resist * res S. Arun-Kumar, CSE
| RES of real Title Page
| SER of resist * res Contents
val value = fn : resist -> real JJ II
val R = PAR (SER (PAR #,SER #),RES 3. J I
val it = () : unit Page 547 of 709
- value R; Go Back
val it = 2.26363636364 : real Full Screen
-
Close
Quit
Resistance Expressions
6.3. User Defined Structured Data Types
1. User Defined Types
IIT Delhi
2. Resistors: Grouping
3. Resistors: In Pairs S. Arun-Kumar, CSE
4. Resistor: Values
Title Page
5. Resistance Expressions
6. Resistance Expressions Contents
7. Arithmetic Expressions
JJ II
8. Arithmetic Expressions: 0
J I
11. Arithmetic Expressions: 3 Page 548 of 709

Go Back
14. Arithmetic Expressions: 6 Full Screen

Close
17. Binary Trees Quit
19. Trees: Traversals
20. Recursive Data Types: Correctness
21. Data Types: Correctness IIT Delhi
S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 549 of 709
Go Back
Full Screen
Close
Quit
User Defined Types IIT Delhi
S. Arun-Kumar, CSE
• Records
Title Page
• Structural Types Contents
– Constructors JJ II
∗ Non-recursive J I
∗ Enumeration Types Page 550 of 709
– Recursive datatypes Go Back
∗ Resistance circuits Full Screen
Close
Quit
Resistors: Grouping
R3 IIT Delhi
5.0 S. Arun-Kumar, CSE
5.0 2.0 Title Page
Contents
4.0 R2 JJ II
R1 J I
Page 551 of 709
3.0 R4 Go Back
Full Screen
+ − Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Resistors: In Pairs Title Page
val R1 = PAR (RES 5.0,RES 4.0) : resi Contents
val R2 = SER (RES 5.0,RES 2.0) : resi

JJ II
val R3 = SER (R1, R2); J I
val R4 = PAR (R3, RES(3.0)); Page 552 of 709
Go Back
Full Screen
Close
Quit
Resistor: Values IIT Delhi
- value R1;
S. Arun-Kumar, CSE
val it = 2.22222222222 : real

Title Page
- value R2;
Contents
val it = 7.0 : real

JJ II
- value R3;
J I
val it = 9.22222222222 : real
Page 553 of 709
- value R4;
- Full Screen
Close
Quit
Resistance
Expressions
A resistance expression IIT Delhi
PAR S. Arun-Kumar, CSE
Title Page
SER Contents
JJ II
J I
PAR SER
Page 554 of 709
Go Back
RES RES RES RES RES Full Screen
Close
5 4 5 2 3
Quit
Circuit Diagram
Resistance
Expressions
A resistance expression
IIT Delhi
R4 PAR S. Arun-Kumar, CSE
Title Page
R3 SER Contents
JJ II
J I
R1 PAR R2 SER
Page 555 of 709
Go Back
RES RES RES RES RES Full Screen
2
Close
5 4 5 3
Quit
Circuit Diagram
IIT Delhi
S. Arun-Kumar, CSE
Arithmetic Expressions Title Page
Contents
ML arithmetic expressions: JJ II
((5 * ˜4) + ˜(5 - 2)) div 3 J I
are represented as trees Page 556 of 709
Go Back
Full Screen
Close
Quit
Arithmetic
Expressions: 0 IIT Delhi
div S. Arun-Kumar, CSE
Title Page
+ Contents
JJ II
J I
* ~ Page 557 of 709
− Go Back
5 ~ 3 Full Screen
Close
4 5 2
Quit
Arithmetic
div S. Arun-Kumar, CSE
Title Page
+ Contents
JJ II
J I
* ~
Page 558 of 709
− Go Back
5 ~ 3 Full Screen
Close
4 5 2
Quit
Arithmetic
S. Arun-Kumar, CSE
div Title Page
Contents
+ JJ II
J I
Page 559 of 709
* ~
Go Back
3 Full Screen
5 −4 3 Close
Quit
Arithmetic
S. Arun-Kumar, CSE
div
Title Page
Contents
+ JJ II
J I
Page 560 of 709
* ~
Go Back
3 Full Screen
5 −4 3 Close
Quit
Arithmetic
S. Arun-Kumar, CSE
div Title Page
Contents
+ JJ II
J I
Page 561 of 709
−20 −3
Go Back
Full Screen
3 Close
Quit
Arithmetic
S. Arun-Kumar, CSE
div Title Page
Contents
+ JJ II
J I
Page 562 of 709
−20 −3
Go Back
Full Screen
3 Close
Quit
Arithmetic
S. Arun-Kumar, CSE
div Title Page
Contents
−23 JJ II
J I
Page 563 of 709
Go Back
Full Screen
3 Close
Quit
Arithmetic
S. Arun-Kumar, CSE
div Title Page
Contents
−23 JJ II
J I
Page 564 of 709
Go Back
Full Screen
3 Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Title Page
Arithmetic Contents
Expressions: 8 JJ II
J I
−8
Page 565 of 709
Go Back
Full Screen
Close
Quit
Binary Trees
IIT Delhi
S. Arun-Kumar, CSE
datatype ’a bintree = Title Page
Empty | Contents
Node of ’a * JJ II
’a bintree * J I
’a bintree Page 566 of 709
Go Back
Full Screen
Close
Quit
Arithmetic
Arithmetic Expressions
S. Arun-Kumar, CSE
div
Title Page
+ Contents
JJ II
J I
* ~
Page 567 of 709
5 ~ − 3 Go Back
Full Screen
4 5 2 Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Trees: Traversals Title Page
Contents
• preorder JJ II
• inorder J I
• postorder Page 568 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Recursive Data Types: Title Page
Contents
Correctness JJ II
Correctness on lists by cases

J I
P is proved by case analysis. Page 569 of 709
Go Back
Full Screen
Close
Quit
Data Types: IIT Delhi
Correctness S. Arun-Kumar, CSE
Title Page
Basis Prove P (c) for each non-

Contents
recursive constructor c
JJ II
Induction hypothesis (IH) Assume J I
P (T ) for all elements of the data Page 570 of 709
type less than a certain depth Go Back
Induction Step Prove P (r(T1, . . . , Tn)) Full Screen
for each recursive constructor r Close
Quit
7. Imperative Programming: An Introduction
7.1. Introducing a Memory Model
1. Summary: Functional Model
IIT Delhi
2. CPU & Memory: Simplified
3. Resource Management S. Arun-Kumar, CSE
4. Shell: User Interface

Title Page
5. GUI: User Interface
6. Memory Model: Simplified

Contents
7. Memory
8. The Imperative Model JJ II
9. State Changes: σ
J I
10. State
11. State Changes Page 571 of 709
12. State Changes: σ

Go Back
13. State Changes: σ1
14. State Changes: σ2 Full Screen
15. Languages
Close
16. User Programs
17. Imperative Languages

Quit
18. Imperative vs Functional Variables
19. Assignment Commands
21. Assignment Commands IIT Delhi

S. Arun-Kumar, CSE
24. Assignment Commands: Swap Title Page
25. Swap
Contents
26. Swap
27. Swap JJ II
J I
Page 572 of 709
Go Back
Full Screen
Close
Quit
Summary: Functional
Model IIT Delhi
• Stateless (as is most mathematics) S. Arun-Kumar, CSE
• Notion of value is paramount Title Page
Contents
– Integers, reals, booleans, strings

JJ II
and characters are all values
J I
– Every function is also a value Page 573 of 709
– Every complex piece of data is Go Back
also a value Full Screen
• No concept of storage (except for Close
space complexity calculations) Quit

CPU & Memory:
Simplified IIT Delhi
S. Arun-Kumar, CSE
Title Page
Memory
Contents
JJ II
Peripherals J I
CPU
Page 574 of 709
Printer
Disk Go Back
Keyboard
Full Screen
Screen Close
Quit
Resource
Management IIT Delhi
S. Arun-Kumar, CSE
Title Page
Memory
Contents
JJ II
Peripherals J I
CPU Printer Page 575 of 709
Disk
Go Back
Keyboard
Screen Full Screen
Close
Operating System
Quit
Shell: User Interface
IIT Delhi
S. Arun-Kumar, CSE
Memory Title Page
Contents
Peripherals JJ II
CPU Printer
J I
Disk
Keyboard Page 576 of 709
Screen
Go Back
Operating System Full Screen
Shell Close
Quit
GUI: User Interface IIT Delhi
S. Arun-Kumar, CSE
Memory Title Page
Contents
Peripherals
JJ II
CPU Printer
Disk
J I
Keyboard
Screen Page 577 of 709
Operating System Go Back
Shell Full Screen
Graphical User Interface (GUI) Close
Quit
Memory Model:
Simplified IIT Delhi
1. A sequence of storage cells S. Arun-Kumar, CSE
2. Each cell is a container of a single Title Page
unit of information. Contents
• integer, real, boolean, character JJ II
or string J I
3. Each cell has a unique name, called Page 578 of 709
its address Go Back
Full Screen
4. The memory cell addresses range
from 0 to (usually) 2k − 1 (for some Close
k) Quit
Memory
0 1 2 3
IIT Delhi
S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 579 of 709
Go Back
Full Screen
Close
32K−1 Quit
The Imperative Model
IIT Delhi
S. Arun-Kumar, CSE
• Memory or Storage made explicit Title Page
• Notion of state (of memory) Contents
JJ II
– State is simply the value con-
J I
tained in each cell.
Page 580 of 709
– state : Addresses → V alues

Go Back
• State changes Full Screen
Close
Quit
State Changes: σ
0 1 2 3
IIT Delhi
S. Arun-Kumar, CSE
4
Title Page
Contents
3.1 JJ II
true
J I
Page 581 of 709
Go Back
"#a" Full Screen
Close
Assume all other cells are filled with null

Quit
IIT Delhi
State S. Arun-Kumar, CSE
Title Page
The state σ
Contents
• σ(12) = 4 : int JJ II
• σ(20) = null J I
• σ(43) = true : bool Page 582 of 709
• σ(66) = ”#a” : char Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
State Changes Title Page
• A state change takes place when Contents
the value in some cell changes JJ II
J I
• The contents of only one cell may
Page 583 of 709
be changed at a time.
Go Back
Full Screen
Close
Quit
State Changes: σ
0 1 2 3
IIT Delhi
S. Arun-Kumar, CSE
4
Title Page
Contents
3.1 JJ II
true
J I
Page 584 of 709
Go Back
"#a" Full Screen
Close

Quit
State Changes: σ1
0 1 2 3
IIT Delhi
5
S. Arun-Kumar, CSE
Title Page
Contents
3.1 JJ II
true
J I
Page 585 of 709
Go Back
"#a" Full Screen
Close

Quit
State Changes: σ2
0 1 2 3
IIT Delhi
5
S. Arun-Kumar, CSE
Title Page
Contents
3.1 JJ II
true
J I
Page 586 of 709
Go Back
12 "#a" Full Screen
Close

Quit
Languages
IIT Delhi
S. Arun-Kumar, CSE
Memory Title Page
Contents
Peripherals JJ II
CPU Printer J I
Disk
Page 587 of 709
Keyboard
Go Back
Screen
Full Screen
Operating System Close
Programming Language Interface Quit

User Programs
IIT Delhi
S. Arun-Kumar, CSE
Memory Title Page
Contents
Peripherals
JJ II
CPU Printer
Disk J I
Keyboard
Page 588 of 709
Screen
Go Back
Operating System
Full Screen
Programming Language Interface Close
User Programs Quit

Imperative Languages
• How is the memory accessed? IIT Delhi
S. Arun-Kumar, CSE
– Through system calls to the OS.

Title Page
• How are memory cells identified? Contents
– Use Imperative variables. JJ II
– Each such variable is a name J I
mapped to an address . Page 589 of 709
• How are state changes accom- Go Back
plished? Full Screen
Close
– By the assignment command.
Quit
Imperative vs IIT Delhi
Functional Variables
S. Arun-Kumar, CSE
Title Page
Functional Imperative Contents
name of a value name of an address JJ II
constant could change with time J I
Page 590 of 709
The value contained in an imperative

Go Back
variable x is denoted !x. Full Screen
Close
Quit
Assignment IIT Delhi
Commands S. Arun-Kumar, CSE
Title Page
Let x and y be imperative variables.

Contents
Consider the following commands.
JJ II
Assuming !x = 1 and !y = 2.
J I
Page 591 of 709
Go Back
x 1 y 2 Full Screen
Close
Quit
Title Page
Store the value 5 in x.

x := 5 Contents
JJ II
J I
Page 592 of 709
x 5 y 2 Go Back
Full Screen
Close
Quit
Title Page
Copy the value contained in y into x.

x :=!y Contents
JJ II
J I
Page 593 of 709
x 2 y 2 Go Back
Full Screen
Close
Quit
Increment the value contained in x by 1. Title Page
x :=!x + 1 Contents
. JJ II
J I
Page 594 of 709
Go Back
x 3 y 2 Full Screen
Close
Quit
Store the product of the values in x and y Title Page
in y. Contents
y :=!x∗!y JJ II
J I
Page 595 of 709
Go Back
x 3 y 6 Full Screen
Close
Quit
Assignment
Commands: Swap IIT Delhi
Swap the values in x and y. S. Arun-Kumar, CSE
Title Page
Contents
Swapping values implies trying to JJ II
make two state changes simultane- J I
ously! Page 596 of 709
Go Back
Full Screen
Requires a new memory cell t to tem- Close
porarily store one of the values. Quit

Swap IIT Delhi
How does one get a new memory S. Arun-Kumar, CSE
cell? Title Page
val t = ref 0
Contents
Then the rest is easy JJ II
J I
val t = ref 0;
t := !x; Page 597 of 709
x := !y; Go Back
y := !t; Full Screen
Close
Quit
IIT Delhi
Swap S. Arun-Kumar, CSE
Title Page
Could be made simpler! Contents
val t = ref (!x); JJ II
x := !y; J I
y := !t; Page 598 of 709
Go Back
Full Screen
Close
Quit
Swap
IIT Delhi
S. Arun-Kumar, CSE
Could use a temporary functional vari- Title Page
able t instead of an imperative vari- Contents
able JJ II
val t = !x; J I
x := !y; Page 599 of 709
y := t; Go Back
Full Screen
Close
Quit
7.2. Imperative Programming:
1. Imperative vs Functional
IIT Delhi
2. Features of the Store
3. References: Experiments S. Arun-Kumar, CSE
4. References: Experiments
Title Page
6. Aliases Contents
JJ II
8. References: Aliases
J I
10. After Garbage Collection
11. Side Effects Page 600 of 709
12. Imperative ML
Go Back
13. Imperative ML
14. Imperative ML Full Screen
15. Imperative ML
Close
16. Nasty Surprises
17. Imperative ML Quit
18. Imperative ML
19. Common Errors
20. Aliasing & References
21. Dangling References IIT Delhi
22. New Reference

S. Arun-Kumar, CSE
23. Imperative Commands: Basic
24. Imperative Commands: Compound Title Page
25. Predefined Compound Commands

Contents
JJ II
J I
Page 601 of 709
Go Back
Full Screen
Close
Quit
Imperative vs IIT Delhi
S. Arun-Kumar, CSE
Functional Title Page
• Functional Model Contents
• Memory/Store Model JJ II
J I
• Imperative Model
Page 602 of 709
• State Changes Go Back
• Accessing the store Full Screen
Close
Quit
Features of the Store
IIT Delhi
S. Arun-Kumar, CSE
Memory is treated as a datatype with Title Page
constructors Contents
Allocation ref : α → α ref JJ II
J I
Dereferencing ! : α ref → α
Page 603 of 709
Updation :=: α ref ∗ α → unit Go Back
Deallocation of memory is automatic! Full Screen
Close
Quit
References:
Experiments
- val a = ref 0; IIT Delhi
val a = ref 0 : int ref S. Arun-Kumar, CSE
Title Page
0 1 2 3
Contents
0
JJ II
J I
Page 604 of 709
Go Back
Full Screen
Close
Quit
a
References:
Experiments
- val b = ref 0; IIT Delhi
val b = ref 0 : int ref S. Arun-Kumar, CSE
Title Page
0 1 2 3
Contents
0
JJ II
J I
0 Page 605 of 709
Go Back
Full Screen
Close
Quit
a b
IIT Delhi
References: S. Arun-Kumar, CSE
Experiments Title Page
Contents
- a = b;
JJ II
val it = false : bool
J I
- !a = !b;
Page 606 of 709

Go Back
Full Screen
Close
Quit
Aliases
- val x = ref 0;
IIT Delhi
val x = ref 0 : int ref

S. Arun-Kumar, CSE
0 1 2 3 Title Page
0
Contents
JJ II
0
J I
0
Page 607 of 709
Go Back
Full Screen
Close
a b x Quit
References:
Experiments
- val y = x; IIT Delhi
val y = ref 0 : int ref S. Arun-Kumar, CSE
Title Page
0 1 2 3
Contents
0
JJ II
0 J I
0 Page 608 of 709
Go Back
Full Screen
Close
Quit
a b x y
References: Aliases
- x := !x + 1;
IIT Delhi
val it = () : unit
S. Arun-Kumar, CSE
0 1 2 3 Title Page
0
Contents
JJ II
1
J I
0
Page 609 of 709
Go Back
Full Screen
Close
a b x y Quit
IIT Delhi
References: S. Arun-Kumar, CSE
Experiments Title Page
Contents
- !y;
JJ II
val it = 1 : int
J I
- x = y;
Page 610 of 709

Go Back
Full Screen
Close
Quit
After Garbage
Collection IIT Delhi
GC #0.0.0.0.2.45: (0 ms) S. Arun-Kumar, CSE
Title Page
0 1 2 3
0 0 1 Contents
JJ II
J I
Page 611 of 709
Go Back
Full Screen
Close
y
Quit
a b x
Side Effects IIT Delhi
• Assignment does not produce a S. Arun-Kumar, CSE
value Title Page
Contents
• It produces only a state change

JJ II
(side effect)
J I
• But side-effects are compatible with Page 612 of 709
functional programming since it is Go Back
provided as a new data type with Full Screen
constructors and destructors.

Close
Quit
Imperative ML IIT Delhi
• Does not provide direct access to S. Arun-Kumar, CSE
memory addresses Title Page
• Does not allow for uninitialized im- Contents
perative variables JJ II
J I
• Provides a type with every memory
Page 613 of 709
location
Go Back
• Manages the memory completely Full Screen
automatically Close
Quit
Imperative ML
• Does not provide direct access to
IIT Delhi
memory addresses
S. Arun-Kumar, CSE
– Prevents the use of memory ad- Title Page
dresses as integers that can be Contents
manipulated by the user program JJ II
• Does not allow for uninitialized im- J I
perative variables Page 614 of 709
• Provides a type with every memory Go Back
location Full Screen
• Manages the memory completely Close
automatically Quit
Imperative ML
IIT Delhi
memory addresses
S. Arun-Kumar, CSE
• Does not allow for uninitialized im- Title Page
perative variables Contents
– Most imperative languages keep JJ II
declarations separate from initial- J I
izations Page 615 of 709
cell Full Screen
automatically Quit
Imperative ML
IIT Delhi
memory addresses
S. Arun-Kumar, CSE
• Does not allow for uninitialized im- Title Page
perative variables Contents
– A frequent source of surprising JJ II
results in most imperative lan- J I
guage programs Page 616 of 709
cell Full Screen
automatically Quit
Nasty Surprises
Separation of declaration from initial-
IIT Delhi
ization
S. Arun-Kumar, CSE
• Uninitialized variables Title Page
• Execution time errors if not de- Contents
tected by compiler, since every JJ II
memory location contains some J I
data Page 617 of 709
• Might use a value stored previously Go Back
in that location by some imperative Full Screen
variable that no longer exists. Close
• Errors due to type violations. Quit

Imperative ML IIT Delhi
• Does not provide direct access to S. Arun-Kumar, CSE
memory addresses Title Page
• Does not allow for uninitialized im- Contents
perative variables JJ II
J I
Page 618 of 709
cell
Go Back
• Manages the memory completely Full Screen
automatically and securely. Close
Quit
Imperative ML
memory addresses IIT Delhi
• Does not allow for uninitialized im- S. Arun-Kumar, CSE
perative variables Title Page
Contents

JJ II
cell
J I
• Manages the memory completely Page 619 of 709
automatically and securely

Go Back
– Memory has to be managed by Full Screen
the user program in most lan- Close
guages Quit
– Prone to various errors

IIT Delhi
Common Errors S. Arun-Kumar, CSE
Title Page
• Memory access errors due to inte- Contents
ger arithmetic, especially in large JJ II
structures (arrays) J I
• Dangling references on dealloca- Page 620 of 709
tion of aliased memory Go Back
Full Screen
Close
Quit
Aliasing & References IIT Delhi
Before deallocation:
0 1 2 3 S. Arun-Kumar, CSE
0 0 1
Title Page
Contents
JJ II
J I
Page 621 of 709
Go Back
Full Screen
a b x y Close
Quit
Dangling References
Deallocate x through a system call
0 1 2 3
IIT Delhi
0 0
S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 622 of 709
Go Back
a b y Full Screen
Close
y is left dangling!
Quit
New Reference
val z = ref 12;
IIT Delhi
0 1 2 3
0 0 12 S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 623 of 709
Go Back
Full Screen
a b y z
Close
Quit
By sheer coincidence !y = 12
IIT Delhi
Imperative Commands: S. Arun-Kumar, CSE
Basic Title Page
A Command is an ML expression that Contents
creates a side effect and returns an JJ II
empty tuple (() : unit). J I
Page 624 of 709
Assignment
Go Back
print Full Screen
Close
Quit
Imperative Commands:
Compound IIT Delhi
S. Arun-Kumar, CSE
Any complex ML expression or func- Title Page
tion definition whose type is of the Contents
form α → unit is a compound com- JJ II
mand. J I
• Predefined ML compound com- Page 625 of 709
mands Go Back
• Could be user-defined. After all, ev- Full Screen
erything is a value! Close
Quit
Predefined Compound
Commands IIT Delhi
branching if e then c1 else c0.

S. Arun-Kumar, CSE
Title Page
cases case e of p1 ⇒ c1| · · · |pn ⇒ cn Contents
Sequencing (c1; c2; . . . ; cn). Sequenc- JJ II
ing is associative J I
looping while e do c1 is defined recur- Page 626 of 709
sively as Go Back
if e then (c1; while e do c1) else () Full Screen
Close
Quit
7.3. Arrays
1. Why Imperative
IIT Delhi
2. Arrays
3. Indexing Arrays S. Arun-Kumar, CSE
4. Indexing Arrays
Title Page
5. Indexing Arrays
6. Physical Addressing Contents
7. Arrays
JJ II
8. 2D Arrays
9. 2D Arrays: Indexing J I
10. Ordering of indices
Page 627 of 709
11. Arrays vs. Lists
12. Arrays: Physical

Go Back
13. Lists: Physical
Full Screen
Close
Quit
Why Imperative
IIT Delhi
• Historical reasons: Early machine

S. Arun-Kumar, CSE
instruction set.
Title Page
• Programming evolved from the ma- Contents
chine architecture.
JJ II
• Legacy software: J I
– numerical packages Page 628 of 709
– operating systems Go Back
• Are there any real benefits of imper- Full Screen
ative programming? Close
Quit
Arrays
An array of length n is a contiguous
sequence of n memory cells IIT Delhi
C0, C1, . . . , Cn−1 S. Arun-Kumar, CSE
Title Page
Contents
JJ II
C0 Ci J I
Cn−1 Page 629 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Indexing Arrays Title Page
For any array Contents
JJ II
• i, 0 ≤ i < n is the index of cell Ci.
J I
• Ci is at a distance of i cells away Page 630 of 709
from C0
Go Back
Full Screen
Close
Quit
Indexing Arrays
IIT Delhi
S. Arun-Kumar, CSE
Title Page
C0 Ci Contents
Cn−1 JJ II
J I
Page 631 of 709
Go Back
Full Screen
a0 a n−1 ai
Close
Quit
IIT Delhi
Indexing Arrays S. Arun-Kumar, CSE
Title Page
• The start address of the array and Contents
the address of C0 are the same (say JJ II

a0)
J I
• The address ai of cell Ci is Page 632 of 709
ai = a0 + i Go Back
Full Screen
Close
Quit
Physical Addressing
If each element occupies s physical IIT Delhi
memory locations, then S. Arun-Kumar, CSE
ai = a0 + i × s
Title Page
Contents
JJ II
J I
Page 633 of 709
Go Back
Full Screen
Close
a0 a n−1 ai Quit
Arrays
IIT Delhi
S. Arun-Kumar, CSE
A 2-dimensional array of Title Page
• r rows numbered 0 to r − 1 Contents
• each row containing c elements JJ II
numbered 0 to c − 1 J I
Page 634 of 709
is also a contiguous sequence of rc

Go Back
memory cells
Full Screen
Close
C0,0, C0,1, . . . , C0,c−1, C1,0, . . . , Cr−1,c−1

Quit
2D Arrays
A 2 dimensional-array is represented IIT Delhi
as an array of length r × c, where S. Arun-Kumar, CSE
• a00 is the start address of the array, Title Page
and Contents
• the address of the (i, j)-th cell is JJ II
given by J I
aij = a00 + (c × i + j) Page 635 of 709
Go Back
• the physical address of the (i, j)-th

Full Screen
cell is given by
Close
aij = a00 + (c × i + j) × s
Quit
2D Arrays: Indexing IIT Delhi
• The index (i, j) of a 2D array may S. Arun-Kumar, CSE
be thought of as being similar to a Title Page
2-digit number in base c Contents
• The successor of index (i, j) is the JJ II
successor of a number in base c i.e. J I
Page 636 of 709

(i + 1, 0) if j = n − 1 Go Back
succ(i, j) =
(i, j + 1) else Full Screen
Close
Quit
IIT Delhi
Ordering of indices S. Arun-Kumar, CSE
Title Page
There is a natural “<” ordering on in-

Contents
dices given by
JJ II
(i, j) < (k, l) ⇐⇒ J I
(i < k) or Page 637 of 709
(i = k and j < l) Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Arrays vs. Lists Title Page
Lists Arrays Contents
JJ II
Unbounded lengths Fixed length
Insertions possible Very complex J I
Indirect access Direct access Page 638 of 709
Go Back
Full Screen
Close
Quit
Arrays: Physical
IIT Delhi
S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 639 of 709
Go Back
Full Screen
Close
a0 a n−1 ai Quit
Lists: Physical
IIT Delhi
S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 640 of 709
Go Back
Full Screen
Close
Quit
8. A large Example: Tautology Checking
8.1. Large Example: Tautology Checking IIT Delhi
1. Logical Arguments
S. Arun-Kumar, CSE
2. Saintly and Rich
3. About Cats Title Page
4. About God
Contents
5. Russell’s Argument
6. Russell’s Argument JJ II
J I
9. Propositions Page 641 of 709
10. Compound Propositions

Go Back
11. Valuations
12. Valuations
Full Screen
13. Tautology
14. Properties Close
15. Negation Normal Form

Quit
16. Literals & Clauses
17. Conjunctive Normal Form
18. Validity
19. Logical Validity IIT Delhi
20. Validity & Tautologies

S. Arun-Kumar, CSE
21. Problem
22. Tautology Checking Title Page
23. Falsifying
Contents
JJ II
J I
Page 642 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
Logical Arguments S. Arun-Kumar, CSE
Title Page
Examples.
Contents
• Saintly and Rich JJ II
• About cats J I
• About God Page 643 of 709
• Russell’s argument Go Back
Full Screen
Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Saintly and Rich Title Page
Contents
hy1 The landed are rich. JJ II
hy2 One cannot be both saintly and rich. J I
conc The landed are not saintly Page 644 of 709
Go Back
Full Screen
Close
Quit
IIT Delhi
About Cats S. Arun-Kumar, CSE
hy1 Tame cats are non-violent and vege- Title Page
tarian. Contents
hy2 Non-violent cats would not kill mice. JJ II
J I
hy3 Vegetarian cats are bottle-fed.
Page 645 of 709
hy4 Cats eat meat. Go Back
conc Cats are not tame. Full Screen
Close
Quit
About God IIT Delhi
S. Arun-Kumar, CSE
hy1 God is omniscient and omnipotent.

Title Page
hy2 An omniscient being would know Contents
there is evil. JJ II
hy3 An omnipotent being would prevent J I
evil. Page 646 of 709
hy4 There is evil. Go Back
conc There is no God Full Screen
Close
Quit
Russell’s Argument
hy1 If we can directly know that God ex- IIT Delhi
ists, then we can know God exists by S. Arun-Kumar, CSE
experience. Title Page
Contents
hy2 If we can indirectly know that God ex-

JJ II
ists, then we can know God exists by
J I
logical inference from experience.
Page 647 of 709
hy3 If we can know that God exists, then Go Back
we can directly know that God exists, Full Screen
or we can indirectly know that God ex-

Close
ists.
Quit
hy4 If we cannot know God empirically, IIT Delhi
then we cannot know God by experi- S. Arun-Kumar, CSE
ence and we cannot know God by logi- Title Page
cal inference from experience. Contents
JJ II
hy5 If we can know God empirically, then
“God exists” is a scientific hypothesis J I
and is empirically justifiable. Page 648 of 709
Go Back
hy6 “God exists” is not empirically justi-

Full Screen
fiable.
Close
conc We cannot know that God exists. Quit

hy1 If we can directly know that God ex- IIT Delhi
ists, then we can know God exists by S. Arun-Kumar, CSE
experience. Title Page
Contents
hy2 If we can indirectly know that God ex-
JJ II
ists, then we can know God exists by
J I
logical inference from experience.
Page 649 of 709
hy3 If we can know that God exists, then

Go Back
(we can directly know that God exists,

Full Screen
or we can indirectly know that God ex-

Close
ists).
Quit
hy4 If we cannot know God empirically, IIT Delhi
then (we cannot know God by experi- S. Arun-Kumar, CSE
ence and we cannot know God by log- Title Page
ical inference from experience.) Contents
hy5 If we can know God empirically, then JJ II
(“God exists” is a scientific hypothesis J I
and is empirically justifiable.) Page 650 of 709
Go Back
hy6 “God exists” is not empirically justi-

Full Screen
fiable.
Close
conc We cannot know that God exists. Quit

Propositions IIT Delhi
S. Arun-Kumar, CSE
A proposition is a sentence to which a Title Page
truth value may be assigned. Contents
In any real or imaginary world of facts JJ II
a proposition has a truth value, true or J I
false. Page 651 of 709
An atom is a simple proposition that Go Back
has no propositions as components. Full Screen
Close
Quit
Compound
Propositions IIT Delhi
Compound propositions may be S. Arun-Kumar, CSE
formed from atoms by using the fol- Title Page
lowing operators/constructors. Contents
JJ II
operator symbol J I
not ¬ Page 652 of 709
and ∧ Go Back
or ∨ Full Screen
if. . . then. . . ⇒ Close
equivalent ⇐⇒ Quit
IIT Delhi
Valuations S. Arun-Kumar, CSE
Given truth values to individual atoms Title Page
the truth values of compound proposi- Contents
tions are evaluated as follows: JJ II
p ¬p J I
true false Page 653 of 709
false true Go Back
Full Screen
Close
Quit
IIT Delhi
Valuations
S. Arun-Kumar, CSE
Title Page
p q p∧q p∨q p⇒q p ⇐⇒ q Contents
true true true true true true JJ II
true false false true false false J I
false true false true true false Page 654 of 709
false false false false true true Go Back
Full Screen
Close
Quit
Tautology IIT Delhi
S. Arun-Kumar, CSE
A (compound) proposition is a tautol-

Title Page
ogy if it is true regardless of what truth

Contents
values are assigned to its atoms.
JJ II
Examples.
J I
• p∨¬p Page 655 of 709
• (p∧q)⇒p Go Back
• (p∧¬p)⇒q Full Screen
Close
Quit
Properties
• Every proposition may be ex- IIT Delhi
pressed in a logically equivalent S. Arun-Kumar, CSE
form using only the operators ¬, ∧ Title Page
and ∨ Contents
(p ⇐⇒ q) = (p⇒q)∧(q⇒p)
JJ II
(p⇒q) = (¬p∨q) J I
• De Morgan’s laws may be applied Page 656 of 709
to push ¬ inward Go Back
¬(p∧q) = ¬p∨¬q Full Screen
Close
¬(p∨q) = ¬p∧¬q
Quit
Negation Normal Form IIT Delhi
S. Arun-Kumar, CSE
• Double negations may be removed Title Page
since Contents
¬¬p = p JJ II
• Every proposition may be ex- J I
pressed in a form containing only ∧ Page 657 of 709
and ∨ with ¬ appearing only in front Go Back
of atoms. Full Screen
Close
Quit
IIT Delhi
Literals & Clauses S. Arun-Kumar, CSE
Title Page
• A literal is either an atom or ¬ ap- Contents
plied to an atom JJ II
• ∨ is commutative and associative J I

Wm
• A clause is of the form j=1lj , Page 658 of 709
where each lj is a literal. Go Back
Full Screen
Close
Quit
Conjunctive Normal IIT Delhi
Form S. Arun-Kumar, CSE
Title Page
• ∨ may be distributed over ∧ Contents
p∨(q∧r) = (p∨q)∧(p∨r) JJ II
• ∧ is commutative and associative. J I
• Every proposition V may be ex- Page 659 of 709
pressed in the form ni=1qi, where Go Back
each qi is a clause. Full Screen
Close
Quit
IIT Delhi
Validity S. Arun-Kumar, CSE
• A logical argument consists of a Title Page
number of hypotheses and a single Contents
conclusion ([h1, . . . , hn]|c) JJ II
J I
• A logical argument is valid if the
Page 660 of 709
conclusion logically follows from the
hypotheses.
Go Back
Full Screen
Close
Quit
Logical Validity IIT Delhi
S. Arun-Kumar, CSE
The conclusion logically follows Title Page
from the given hypotheses if for Contents
any truth assignment to the atoms,

JJ II
either some hypothesis hi is false J I
or whenever every one of the hypothe- Page 661 of 709
ses is true the conclusion is also Go Back
true Full Screen
Close
Quit
Validity & Tautologies IIT Delhi
S. Arun-Kumar, CSE
• A tautology is a valid argument

Title Page
in which there is a conclusion
without any hypothesis.
Contents
JJ II
• A logical argument [h1, . . . , hn]|c, is J I
valid if and only if
Page 662 of 709
(h1∧ . . . ∧hn)⇒c Go Back
is a tautology Full Screen
Close
Quit
IIT Delhi
Problem S. Arun-Kumar, CSE
Given an argument [h1, . . . , hn]|c, Title Page
Contents
• determine whether (h1∧ . . . ∧hn)⇒c

JJ II
is a tautology, and
J I
• If it is not a tautology, to deter- Page 663 of 709
mine what truth assignments to the Go Back
atoms make it false. Full Screen
Close
Quit
Tautology Checking IIT Delhi
Vn
A proposition in CNF ( i=1pi) S. Arun-Kumar, CSE
• is a tautology if and only if every Title Page
proposition pi , 1 ≤ i ≤ m, is a tau- Contents
tology. JJ II
• otherwise at least one clause pi J I
must be false Page 664 of 709
Wm
• Clause pi = j=1lij is false
Go Back
if and only if every literal lij , 1 ≤ j ≤ Full Screen
m is false Close
Quit
Falsifying IIT Delhi
Vn
For a proposition in CNF ( i=1pi) that S. Arun-Kumar, CSE
is not a tautology Title Page
Wm
• A clause pi = j=1lij is false Contents
JJ II
• A truth assignment that falsifies the
J I
argument
Page 665 of 709
– sets the atoms that occur nega- Go Back
tively in pi to true,
Full Screen
– sets every other atom to false Close
Quit
8.2. Tautology Checking Contd.
1. Tautology Checking
IIT Delhi
2. Normal Forms
3. Top-down Development S. Arun-Kumar, CSE
4. The Signature
Title Page
5. The Core subproblem
6. The datatype Contents
7. Convert to CNF
JJ II
8. Rewrite into NNF
9. ⇐⇒ and ⇒ Elimination J I
10. Push ¬ inward
Page 666 of 709
11. Push ¬ inward
12. conj of disj

Go Back
13. Push ∨ inward
14. Tautology & Falsification Full Screen
Close
Quit
Tautology Checking IIT Delhi
S. Arun-Kumar, CSE
• Logical arguments Title Page
• Propositional forms Contents
• Propositions JJ II
J I
• Compound Propositions
Page 667 of 709
• Truth table Go Back
• Tautologies Full Screen
Close
Quit
Normal Forms
IIT Delhi
S. Arun-Kumar, CSE
• Properties Title Page
• Negation Normal Form Contents
JJ II
• Conjunctive Normal Forms
J I
• Valid Propositional Arguments as Page 668 of 709
tautologies
Go Back
• The problem Full Screen
Close
Quit
Top-down
Development IIT Delhi
• Transform the argument into a S. Arun-Kumar, CSE
single proposition. Title Page
• Transfom the single proposition into Contents
one in CNF JJ II
J I
• Check whether every clause is a Page 669 of 709
tautology Go Back
Full Screen
• If any clause is not a tautology, find

Close
the truth assignment(s) that make it

Quit
false
The Signature
signature PropLogic = IIT Delhi
sig datatype Prop = ?? S. Arun-Kumar, CSE
type Argument = Title Page
Prop list * Prop Contents
val falsifyArg : JJ II
Argument -> Prop list list
J I
val Valid:
Page 670 of 709
Argument -> bool *

Go Back
Prop list list

Full Screen
...
Close
end;
Quit
The Core subproblem IIT Delhi
S. Arun-Kumar, CSE
• Representing propositions Title Page
• Transformation of propositions into Contents
CNF JJ II
– Transform into Negation Normal J I
Form (NNF) Page 671 of 709
– Transform NNF into Conjunctive Go Back
Normal Form (CNF) Full Screen
Close
Quit
The datatype IIT Delhi
datatype Prop = S. Arun-Kumar, CSE
ATOM of string | Title Page
NOT of Prop | Contents
AND of Prop * Prop | JJ II
OR of Prop * Prop | J I
IMP of Prop * Prop | Page 672 of 709
EQL of Prop * Prop Go Back
Full Screen
Close
Quit
Convert to CNF
IIT Delhi
Convert a given proposition into CNF

S. Arun-Kumar, CSE
fun cnf (P) =

Title Page
conj_of_disj (
Contents
nnf (rewrite (P)));

JJ II
J I
where
Page 673 of 709
• rewrite eliminates ⇐⇒ and ⇒ Go Back
• nnf converts into NNF Full Screen
• conj of disj converts into CNF Close
Quit
IIT Delhi
S. Arun-Kumar, CSE
Rewrite into NNF Title Page
• Eliminate ⇐⇒ and then ⇒ Contents
JJ II
• Push ¬ inward using De
J I
Morgan’s laws and eliminate
double negations. Page 674 of 709
Go Back
Full Screen
Close
Quit
⇐⇒ and ⇒ Elimination IIT Delhi
fun rewrite (ATOM a) = ATOM a S. Arun-Kumar, CSE
| rewrite (IMP (P, Q)) = Title Page
OR (NOT (rewrite(P)), Contents
rewrite(Q))
JJ II
| rewrite (EQL (P, Q)) =
J I
rewrite (AND (IMP(P, Q),
Page 675 of 709
IMP (Q, P)))

Go Back
| ...
Full Screen
Proposition made up of only ¬, ∧ and

∨. Close
Quit
IIT Delhi
Push ¬ inward S. Arun-Kumar, CSE
fun nnf (ATOM a) = Title Page
ATOM a Contents
| nnf (NOT (ATOM a)) = JJ II
NOT (ATOM a) J I
| nnf (NOT (NOT (P))) = Page 676 of 709
nnf (P) Go Back
Full Screen
Close
Quit
Push ¬ inward IIT Delhi
| nnf (NOT (AND (P, Q))) = S. Arun-Kumar, CSE
nnf (OR (NOT (P), Title Page
NOT (Q))) Contents
| nnf (NOT (OR (P, Q))) = JJ II
nnf (AND (NOT (P), J I
NOT (Q))) Page 677 of 709
| ... Go Back
Proposition made up of only ∧ and ∨ Full Screen
applied to positive or negative literals. Close
Quit
conj of disj IIT Delhi
fun conj_of_disj (AND (P, Q)) = S. Arun-Kumar, CSE
AND (conj_of_disj (P), Title Page
conj_of_disj (Q)) Contents
| conj_of_disj (OR (P, Q)) = JJ II
distOR (conj_of_disj (P), J I
conj_of_disj (Q)) Page 678 of 709
| conj_of_disj (P) = P Go Back
Full Screen
where distOR is Close
Quit
Push ∨ inward IIT Delhi
S. Arun-Kumar, CSE
Use distributivity of ∨ over ∧

Title Page
fun distOR (P, AND (Q, R)) =

Contents
AND (distOR (P, Q),

JJ II
distOR (P, R))
J I
| distOR (AND (Q, R), P) =
AND (distOR (Q, P), Page 679 of 709
distOR (R, P)) Go Back
| distOR (P, Q)= OR (P, Q) Full Screen
Close
Quit
Tautology &
Falsification IIT Delhi
Falsifying a proposition S. Arun-Kumar, CSE
• A proposition Q in CNF, not a tau- Title Page
tology if and only if at least one of Contents
the clauses can be made false, by JJ II
a suitable truth assignment J I
• The list of atoms which are set true Page 680 of 709
to falsify a clause is called a falsifier. Go Back
Full Screen
• A proposition is a tautology
Close
if and only if there is no falsifier!

Quit
9. Lecture-wise Index to Slides
IIT Delhi
Index
Introduction to Computing (3-8) S. Arun-Kumar, CSE
1. Introduction
Title Page
2. Computing tools
3. Ruler and Compass Contents
4. Computing and Computers
5. Primitives JJ II
6. Algorithm
7. Problem: Doubling a Square J I
8. Solution: Doubling a Square

Page 681 of 709
10. Execution: Step 2 Go Back
11. Doubling a Square: Justified
12. Refinement: Square Construction Full Screen
13. Refinement 2: Perpendicular at a point

Close
14. Solution: Perpendicular at a point
15. Perpendicular at a point: Justification
Quit
Next: Our Computing Tool
Our Computing Tool (9-33)
Previous: Introduction to Computing IIT Delhi
1. The Digital Computer: Our Computing Tool

S. Arun-Kumar, CSE
2. Algorithms
3. Programming Language
Title Page
4. Programs and Languages
5. Programs Contents
6. Programming
7. Computing Models JJ II
8. Primitives
J I
9. Primitive expressions
10. Methods of combination Page 682 of 709
11. Methods of abstraction
12. The Functional Model Go Back

Full Screen
Close
16. A Functional Program: Factorial
17. A Computation: Factorial Quit
21. A Computation: Factorial IIT Delhi
23. A Computation: Factorial S. Arun-Kumar, CSE
24. Standard ML
25. SML: Important Features Title Page
Next: Primitives: Integer & Real Contents
Primitives: Integer & Real (34-60) JJ II
Previous: Primitives: Integer & Real

J I
1. Algorithms & Programs
2. SML: Primitive Integer Operations 1 Page 683 of 709

Go Back
Full Screen
7. SML: Primitive Integer Operations 1 Close
9. SML: Primitive Integer Operations 2 Quit
13. SML: Primitive Integer Operations 3 IIT Delhi
15. SML: Primitive Integer Operations 3 S. Arun-Kumar, CSE

Title Page
18. Quotient & Remainder
Contents
20. SML: Primitive Real Operations 1 JJ II
22. SML: Primitive Real Operations 2 J I

Page 684 of 709
25. SML: Precision
Go Back
26. Fibonacci Numbers
27. Euclidean Algorithm Full Screen
Next: Technical Completeness & Algorithms Close
Technical Completeness & Algorithms (61-92)

Quit
1. Recapitulation: Integers & Real
2. Recap: Integer Operations
3. Recapitulation: Real Operations
4. Recapitulation: Simple Algorithms
IIT Delhi
5. More Algorithms
6. Powering: Math S. Arun-Kumar, CSE
7. Powering: SML
8. Technical completeness Title Page
9. What SML says

Contents
11. What SML says ... contd
JJ II
12. Powering: Math 1
13. Powering: SML 1 J I
14. Technical Completness
15. What SML says Page 685 of 709
16. Powering: Integer Version

Go Back
17. Exceptions: A new primitive
18. Integer Power: SML
Full Screen
19. Integer Square Root 1
20. Integer Square Root 2 Close
21. An analysis
22. Algorithmic idea Quit
23. Algorithm: isqrt
24. Algorithm: shrink
25. SML: shrink
26. SML: intsqrt IIT Delhi
27. Run it!
28. SML: Reorganizing Code S. Arun-Kumar, CSE

Title Page
30. shrink: Another algorithm
31. Shrink2: SML
Contents
32. Shrink2: SML ... contd
Algorithm Refinement (93-122) JJ II

J I
2. Recap: Power
3. Recap: Technical completeness Page 686 of 709

5. Intsqrt: Reorganized Go Back
Full Screen
7. Some More Algorithms
8. Combinations: Math
Close
9. Combinations: Details
10. Combinations: SML Quit
11. Perfect Numbers
12. Refinement
13. Perfect Numbers: SML
Pu
14. l if divisor(k) IIT Delhi
15. SML: sum divisors
16. if divisor and ifdivisor S. Arun-Kumar, CSE
17. SML: Assembly 1

Title Page
18. SML: Assembly 2
Contents
21. SML: Assembly 3 JJ II
23. Perfect Numbers: Run J I

Page 687 of 709
Go Back
27. Summation: Generalizations
28. Algorithmic Improvements: Full Screen
29. Algorithmic Variations

30. Algorithmic Variations Close
Variations: Algorithms & Code (123-149) Quit

1. Recap
2. Recap: Combinations
3. Combinations 1
4. Combinations 2
IIT Delhi
5. Combinations 3
6. Perfect 2 S. Arun-Kumar, CSE
7. Power 2
8. A Faster Power Title Page
9. Power2: Complete
Contents
10. Power2: SML
11. Power2: SML
JJ II
12. Computation: Issues
13. General Correctness J I
14. Code: Justification
15. Recall Page 688 of 709
16. Features: Definition before Use

Go Back
17. Run ifdivisor
18. Diagnosis: Features of programs
Full Screen
19. Back to Math
20. Incorrectness Close
21. ifdivisor3
22. Run it! Quit
23. Try it!
24. Hey! Wait a minute!
25. The n’s
26. Scope IIT Delhi
27. Scope Rules
S. Arun-Kumar, CSE
Names, Scopes & Recursion (150-181)
1. Disjoint Scopes Title Page
2. Nested Scopes
3. Overlapping Scopes Contents
4. Spannning
5. Scope & Names JJ II

J I
8. Names & References Page 689 of 709

10. Names & References Go Back

Full Screen
Close
15. Names & References Quit
16. Definition of Names
17. Use of Names
19. local...in...end IIT Delhi
21. local...in...end S. Arun-Kumar, CSE
22. Scope & local

Title Page
23. Computations: Simple
24. Simple computations
Contents
26. Composition: Alternative JJ II
27. Compositions: Compare
28. Compositions: Compare J I

Page 690 of 709
30. Recursion
31. Recursion: Left
Go Back
32. Recursion: Right
Floating Point (182-201) Full Screen
1. So Far-1: Computing
Close
2. So Far-2: Algorithms & Programs
3. So far-3: Top-down Design Quit
4. So Far-4: Algorithms to Programs
5. So far-5: Caveats
6. So Far-6: Algorithmic Variations
7. So Far-7: Computations IIT Delhi
8. Floating Point
9. Real Operations S. Arun-Kumar, CSE
10. Real Arithmetic

Title Page
11. Numerical Methods
12. Errors
Contents
13. Errors
14. Infinite Series JJ II
15. Truncation Errors
16. Equation Solving J I
17. Root Finding-1

Page 691 of 709
18. Root Finding-2
19. Root Finding-3
Go Back
20. Root Finding-4
Root Finding, Composition and Recursion (202-229) Full Screen

Close
3. Root Finding: Newton’s Method Quit
IIT Delhi
9. Newton’ Method: Algorithm S. Arun-Kumar, CSE

11. What can go wrong!-2 Title Page

Contents
JJ II
16. Real Computations & Induction: 2 J I
17. What’s it good for? 1
18. What’s it good for? 2 Page 692 of 709
19. newton: Computation

Go Back
20. Generalized Composition
21. Two Computations of h(1)
Full Screen
22. Two Computations of h(−1)
23. Recursive Computations Close
24. Recursion: Left
25. Recursion: Right Quit
26. Recursion: Nonlinear
Termination and Space Complexity (230-266) IIT Delhi
1. Recursion Revisited
S. Arun-Kumar, CSE
2. Linear Recursion: Waxing
3. Recursion: Waning Title Page
4. Nonlinear Recursions
5. Fibonacci: contd Contents
6. Recursion: Waxing & Waning

7. Unfolding Recursion JJ II
8. Non-termination
J I
9. Termination
10. Proofs of termination Page 693 of 709
11. Proofs of termination: Induction

12. Proof of termination: Factorial Go Back

Full Screen
14. Fibonacci: Termination
15. GCD computations
Close
16. Well-foundedness: GCD
17. Well-foundedness Quit
20. Where it doesn’t work
21. Well-foundedness is inductive IIT Delhi
22. Well-foundedness is inductive
23. GCD: Well-foundedness S. Arun-Kumar, CSE

Title Page
26. Example: Zero
Contents
27. Questions
28. The Collatz Problem JJ II
29. Questions
30. Space Complexity J I
31. Newton & Euclid: Absolute

Page 694 of 709
32. Newton & Euclid: Relative
33. Deriving space requirements
Go Back
34. GCD: Space
35. Factorial: Space Full Screen
36. Fibonacci: Space

37. Fibonacci: Space Close
Efficiency Measures and Speed-ups (267-295) Quit

1. Recapitulation
2. Recapitulation
3. Time & Space Complexity
4. isqrt: Space
IIT Delhi
5. Time Complexity
6. isqrt: Time Complexity S. Arun-Kumar, CSE
7. isqrt2: Time
8. shrink vs. shrink2: Times Title Page
9. Factorial: Time Complexity

Contents
10. Fibonacci: Time Complexity
11. Comparative Complexity
JJ II
12. Comparisons
13. Comparisons J I
14. Efficiency Measures: Time
15. Efficiency Measures: Space Page 695 of 709
16. Speeding Up: 1

Go Back
17. Speeding Up: 2
18. Factoring out calculations
Full Screen
20. Tail Recursion: 2 Close
21. Factorial: Tail Recursion
22. Factorial: Tail Recursion Quit
23. A Computation
24. Factorial: Issues
26. Fibonacci: Tail Recursion IIT Delhi
27. fibTR: SML
28. State in Tail Recursion S. Arun-Kumar, CSE
29. Invariance
Title Page
Invariance & Correctness (296-319)
1. Recap Contents
2. Recursion Transformation
3. Tail Recursion: Examples JJ II
4. Comparisons
J I
5. Transformation Issues
6. Correctness Issues 1 Page 696 of 709
7. Correctness Issues 2
8. Correctness Theorem Go Back

Full Screen
11. Invariance Lemma: f actL tr
Close
13. Invariance: Example Quit
14. Proof
15. Invariance Lemma: f ib iter
16. Proof
17. Correctness: Fibonacci IIT Delhi
19. Variants & Invariants S. Arun-Kumar, CSE
20. More Invariants

Title Page
21. Fast Powering 1
22. Fast Powering 2
Contents
23. Root Finding: Bisection
24. Advantage Bisection JJ II
Tuples, Lists & the Generation of Primes (320-341)
J I
1. Recap: Tail Recursion
2. Examples: Invariants Page 697 of 709
3. Tuples
4. Lists Go Back
5. New Lists
Full Screen
6. List Operations
7. List Operations: cons
Close
8. Generating Primes upto n
9. More Properties Quit
10. Composites
11. Odd Primes
12. primesU pto(n)
13. generateF rom(P, m, n, k) IIT Delhi
14. generateF rom
15. primeW RT (m, P ) S. Arun-Kumar, CSE

Title Page
17. primeW RT
18. Density of Primes
Contents
20. The Prime Number Theorem JJ II
21. Complexity
22. Diagnosis J I
Compound Data & Lists (342-367)

Page 698 of 709
1. Compound Data
2. Recap: Tuples Go Back
3. Tuple: Formation
Full Screen
4. Tuples: Selection
5. Tuples: Equality
Close
6. Tuples: Equality errors
7. Lists: Recap Quit
8. Lists: Append
9. cons vs. @
10. Lists of Functions
11. Lists of Functions IIT Delhi
12. Arithmetic Sequences
13. Tail Recursion S. Arun-Kumar, CSE
14. Tail Recursion Invariant

Title Page
15. Tail Recursion
16. Another Tail Recursion: AS3
Contents
17. Another Tail Recursion: AS3 iter
18. AS3: Complexity JJ II
19. Generating Primes: 2
20. primes2U pto(n) J I
21. generate2F rom(P, m, n, k)

Page 699 of 709
22. generate2F rom
23. prime2W RT (m, P )
Go Back
24. prime2W RT
25. primes2: Complexity Full Screen
26. primes2: Diagnosis

Close
Compound Data & List Algorithms (368-394)
1. Compound Data: Summary Quit
2. Records: Constructors
5. Records: Destructors
IIT Delhi
6. Records: Equality
7. Tuples & Records S. Arun-Kumar, CSE
8. Back to Lists
9. Lists: Correctness Title Page
10. Lists: Case Analysis

Contents
11. Lists: Correctness by Cases
12. List-functions: length
JJ II
13. List Functions: search
14. List Functions: search2 J I
15. List Functions: ordered
16. List Functions:insert Page 700 of 709
17. List Functions: reverse

Go Back
18. List Functions: reverse2
Full Screen
21. List Functions:merge contd. Close
22. ML: merge
23. Sorting by Insertion Quit
26. Functions as Data
27. Higher Order Functions IIT Delhi
Higher Order Functions (395-419)

S. Arun-Kumar, CSE
1. Summary: Compound Data
2. List: Examples Title Page
3. Lists: Sorting
4. Higher Order Functions Contents
5. An Example
6. Currying JJ II
7. Currying: Contd
J I
8. Generalization
9. Generalization: 2 Page 701 of 709
10. Applying a list

11. Trying it out Go Back
12. Associativity
Full Screen
13. Apply to a list
14. Sequences
Close
16. Further Generalization Quit
17. Sequences
18. Efficient Generalization
19. Sequences: 2
20. More Generalizations IIT Delhi
21. More Summations
22. Or Maybe . . . Products S. Arun-Kumar, CSE
N
23. Or Some Other
Title Page
N
24. Other
25. Examples of ⊗, e
Contents
Structured Data (420-451)
1. Transpose of a Matrix JJ II
2. Transpose: 0
J I
3. Transpose: 10
4. Transpose: 01 Page 702 of 709
5. Transpose: 20
6. Transpose: 02 Go Back
7. Transpose: 30
Full Screen
8. Transpose: 03
9. trans
Close
10. is2DM atrix
11. User Defined Types Quit
13. User Defined Structural Types
14. Functions vs. data
15. Data as 0-ary Functions IIT Delhi
16. Data vs. Functions
17. Data vs. Functions: Recursion S. Arun-Kumar, CSE
18. Lists
Title Page
19. Constructors
20. Shapes
Contents
21. Shapes: Triangle Inequality
22. Shapes: Area JJ II
23. Shapes: Area
24. ML: Try out J I
25. ML: Try out (contd.)

Page 703 of 709
27. Recursive Data Types
Go Back
28. Resistors: Datatype
29. Resistors: Equivalent Full Screen
30. Resistors
31. Resistors: Example Close
32. Resistors: ML session

Quit
User Defined Structured Data Types (452-472)
1. User Defined Types
2. Resistors: Grouping
3. Resistors: In Pairs IIT Delhi
4. Resistor: Values
S. Arun-Kumar, CSE
Title Page
7. Arithmetic Expressions
8. Arithmetic Expressions: 0 Contents
10. Arithmetic Expressions: 2 JJ II
J I
Page 704 of 709
15. Arithmetic Expressions: 7 Go Back

17. Binary Trees Full Screen

Close
19. Trees: Traversals
20. Recursive Data Types: Correctness Quit
21. Data Types: Correctness
Introducing a Memory Model (473-499)
1. Summary: Functional Model
2. CPU & Memory: Simplified IIT Delhi
3. Resource Management
S. Arun-Kumar, CSE
4. Shell: User Interface
5. GUI: User Interface Title Page
6. Memory Model: Simplified
7. Memory Contents
8. The Imperative Model

9. State Changes: σ JJ II
10. State
J I
11. State Changes
12. State Changes: σ Page 705 of 709
13. State Changes: σ1

14. State Changes: σ2 Go Back
15. Languages
Full Screen
16. User Programs
17. Imperative Languages
Close
18. Imperative vs Functional Variables
19. Assignment Commands Quit
23. Assignment Commands IIT Delhi
24. Assignment Commands: Swap
25. Swap S. Arun-Kumar, CSE
26. Swap
Title Page
27. Swap
Imperative Programming: (500-524) Contents
1. Imperative vs Functional
2. Features of the Store JJ II
J I
5. References: Experiments Page 706 of 709
6. Aliases
7. References: Experiments Go Back
8. References: Aliases
Full Screen
10. After Garbage Collection
Close
11. Side Effects
12. Imperative ML Quit
13. Imperative ML
14. Imperative ML
15. Imperative ML
16. Nasty Surprises IIT Delhi
17. Imperative ML
18. Imperative ML S. Arun-Kumar, CSE
19. Common Errors

Title Page
20. Aliasing & References
21. Dangling References
Contents
22. New Reference
23. Imperative Commands: Basic JJ II
24. Imperative Commands: Compound
25. Predefined Compound Commands J I
Arrays (525-537)
Page 707 of 709
1. Why Imperative
2. Arrays Go Back
3. Indexing Arrays
Full Screen
4. Indexing Arrays
5. Indexing Arrays
Close
6. Physical Addressing
7. Arrays Quit
8. 2D Arrays
9. 2D Arrays: Indexing
10. Ordering of indices
11. Arrays vs. Lists IIT Delhi
12. Arrays: Physical
13. Lists: Physical S. Arun-Kumar, CSE
Large Example: Tautology Checking (538-560)

Title Page
1. Logical Arguments
2. Saintly and Rich Contents
3. About Cats
4. About God JJ II
J I
7. Russell’s Argument Page 708 of 709
9. Propositions Go Back
10. Compound Propositions

Full Screen
11. Valuations
12. Valuations
Close
13. Tautology
14. Properties Quit
15. Negation Normal Form
16. Literals & Clauses
17. Conjunctive Normal Form
18. Validity IIT Delhi
19. Logical Validity
20. Validity & Tautologies S. Arun-Kumar, CSE
21. Problem
Title Page
22. Tautology Checking
23. Falsifying
Contents
Tautology Checking Contd. (561-574)
1. Tautology Checking JJ II
2. Normal Forms
J I
3. Top-down Development
4. The Signature Page 709 of 709
5. The Core subproblem

6. The datatype Go Back
7. Convert to CNF
Full Screen
8. Rewrite into NNF
9. ⇐⇒ and ⇒ Elimination
Close
10. Push ¬ inward
11. Push ¬ inward Quit
12. conj of disj
13. Push ∨ inward
14. Tautology & Falsification
IIT Delhi
S. Arun-Kumar, CSE
Title Page
Contents
JJ II
J I
Page 710 of 709
Go Back
Full Screen
Close
Quit

The Basic Principles To Programming

Uploaded by

Document Information

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

The Basic Principles To Programming

Uploaded by

Copyright:

Available Formats

IIT Delhi

Introduction to Title Page

sak@cse.iitd.ernet.in Page 1 of 709

Department of Computer Science and Engineering Go Back

I. I. T. Delhi, Hauz Khas, New Delhi 110 016.

March 19, 2007 Close

2 Algorithms: Design & Refinement 114

3 Introducing Reals 242

5 Compound Data 390 Go Back

5.1 Tuples, Lists & the Generation of Primes . . . . . . . . . . . . . . . . . . . . . . . 390

6 Higher Order Functions & Structured Data 486 Close

7 Imperative Programming: An Introduction 570

8 A large Example: Tautology Checking 640

3. Ruler and Compass S. Arun-Kumar, CSE

4. Computing and Computers

8. Solution: Doubling a Square JJ II

11. Doubling a Square: Justified Page 4 of 709

12. Refinement: Square Construction

14. Solution: Perpendicular at a point Full Screen

15. Perpendicular at a point: Justification

• This course is about computing Title Page

• Computing as a process is nearly Contents

• Computing as a mental process J I

• Computing may be done with a va- Page 5 of 709

riety of tools which may or may not Go Back

assist the mind Full Screen

• Sticks and stones (counting) Title Page

• Paper and pencil (an aid to mental Contents

• Abacus (still used in Japan!) J I

• Slide rules (ask a retired engineer!)

• Ruler and compass Full Screen

Actually it is a computing tool!

• Construct a length that is half of a Contents

• Construct a square that is twice the Page 7 of 709

area of a given square Go Back

Computers S. Arun-Kumar, CSE

• Computing is much more funda- Contents

• Computing may be done without a J I

computer too! Page 8 of 709

• But a Computer cannot do much Go Back

besides computing. Full Screen

called primitive operations or primi-

Ruler: Can specify lengths, lines Contents

Compass: Can define arcs and cir- JJ II

• The primitives may be combined in Page 9 of 709

various ways to perform a computa- Go Back

tion. Full Screen

• Example Constructing a right bisec- Close

tor of a given line segment. Quit

Algorithm S. Arun-Kumar, CSE

Given a problem to be solved with a Title Page

given tool, the attempt is to evolve a Contents

combination of primitives of the tool in JJ II

a certain order to solve the problem. J I

An explicit statement of this combina- Page 10 of 709

tion along with the order is an algo- Go Back

rithm Full Screen

Problem: Doubling a S. Arun-Kumar, CSE

Square Title Page

Given a square, construct another Contents

square of twice the area of the origi- JJ II

Solution: Doubling a S. Arun-Kumar, CSE

Square Title Page

2. Complete the square ACEF on Go Back