Scribd
Upload
30 views

CS109A Notes For Lecture 1/24/96 Proving Recursive Programs Work

1) The lecture discusses proving the correctness of recursive programs through induction on the number of recursive calls. 2) As an example, it proves a recursive binary converter program correctly outputs the binary representation of an integer by induction on the input value. 3) It also covers the mergesort algorithm, which sorts a list recursively by splitting it into two halves, sorting each half, and merging the results.

Uploaded by

savio77
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
30 views

CS109A Notes For Lecture 1/24/96 Proving Recursive Programs Work

1) The lecture discusses proving the correctness of recursive programs through induction on the number of recursive calls. 2) As an example, it proves a recursive binary converter program correctly outputs the binary representation of an integer by induction on the input value. 3) It also covers the mergesort algorithm, which sorts a list recursively by splitting it into two halves, sorting each half, and merging the results.

Uploaded by

savio77
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

CS109A Notes for Lecture 1/24/96

Proving Recursive Programs Work


When a program is recursive, we can often
nd a natural inductive proof that it works.
The induction is often on the number of
recursive calls that must be made by a
given call to the recursive function | or
some equivalent parameter.
Example: Let's consider our recursive binaryconverter:
(1)
(2)
(3)

void convert(int i) {
if(i>0) {
convert(i/2);
putchar('0' + i%2);
}
}

Statement to be proved:
S (i): convert produces the binary representation of integer i 0.
Note that is the correct binary representation of 0 in this context.
Basis: i = 0. The test of line (1) fails, so is
printed.
Induction: Assume S (j ) for 0 j < i and prove
S (i) for i
0.
Note, we are breaking our habit of proving S (i + 1) from smaller cases; but it
doesn't matter whether we call the next
case to be proven i + 1 or i.
If i is even, say i = 2j , then convert prints
the binary representation of j at line (2) followed by 0 at line (3).
Appending the nal 0 multiplies the
value printed by 2, which gives the representation of 2j , or i.
1

If i is odd, say i = 2j + 1, then line (2) again


prints the binary representation of j (because
i/2 throws away the remainder), and line (3)
prints 1.
Appending 1 has the e ect of multiplying
the value printed by 2 and then adding
1. Again, the result is i, since i = 2j + 1
in this case.

Sorting

To sort a list = to reorder its elements so each


precedes the next according to some ordering .
That is, (a1 ; a2; : : : ; an) is sorted if
a1

a2

an

a3

Mergesort

A recursive sorting algorithm:


Basis: A list of length 1 is already sorted.
Induction: For lists of 1 element
1. Split the list into two equal-as-possible parts.
2. Recursively sort each part.
3. Merge the results by repeatedly pulling the
smaller element from the fronts of the two
sorted lists.
Mergesort is the preferred method for external sorting (sorting of lists so large that disks
must be used).
Example: (3; 1; 4; 1; 5; 9; 2; 6).
1. Split, say (3; 1; 4; 1) and (5; 9; 2; 6).
2. Sort recursively (details omitted): (1; 1; 3; 4)
and (2; 5; 6; 9).
3. Merge: (1; 1; 2; 3; 4; 5; 6; 9).

Splitting

We could split in many ways, e.g.


2

1. Count the list; say n elements. Put the rst


n=2 in one sublist and the remainder in the
other.
2. Iteratively \deal" the elements to the two sublists, keeping track of which list gets the next
element.
3. Both FCS and EMLP give a recursive algorithm that deals the elements two at a time.
This approach avoids having to remember the
\state" (whose turn it is).
Basis: If there are 0 elements, do nothing. If
there is 1 element, give it to the rst sublist.
Induction: If there are n 2 elements, deal one
to each sublist and recursively split the remaining
n
2 elements.

Merging
Basis: If one list is empty, the other list is the

sorted result.

Induction: If neither list is empty, pick the

smaller of the head elements. The result is the


selected element followed by the result of merging
the remaining lists.

Example:

List1
(1; 1; 3; 4)
(1; 3; 4)
(3; 4)
(3; 4)
(4)

List2
(2; 5; 6; 9)
(2; 5; 6; 9)
(2; 5; 6; 9)
(5; 6; 9)
(5; 6; 9)
(5; 6; 9)

Result
(1)
(1; 1)
(1; 1; 2)
(1; 1; 2; 3)
(1; 1; 2; 3; 4)
(1; 1; 2; 3; 4; 5; 6; 9)

You might also like