Problem 7.1

Hoare partition correctness

The version of PARTITION given in this chapter is not the original partitioning algorithm. Here is the original partition algorithm, which is due to C.A.R. Hoare:

HOARE-PARTITION(A, p, r)
x = A[p]
i = p - 1
j = r + 1
while TRUE
repeat
j = j - 1
until A[j] ≤ x
repeat
i = i + 1
until A[i] ≥ x
if i < j
exchange A[i] with A[j]
else return j

1. Demonstrate the operation of HOARE-PARTITION on the array $A = \langle 13, 19, 9, 5, 12, 8, 7, 4, 11, 2, 6, 21 \rangle$, showing the values of the array and auxiliary values after each iteration of the while loop in lines 4-13.

The next three questions ask you to give a careful argument that the procedure HOARE-PARTITION is correct. Assuming that the subarray $A[p..r]$ contains at least two elements, prove the following:

1. The indices $i$ and $j$ are such that we never access an element of $A$ outside the subarray $A[p..r]$.
2. When HOARE-PARTITION terminates, it returns a value $j$ such that $p \le j < r$.
3. Every element of $A[p..j]$ is less than or equal to every element of $A[j + 1..r]$ when HOARE-PARTITION terminates.

The PARTITION procedure in section 7.1 separates the pivot value (originally in $A[r]$) from the two partitions it forms. The HOARE-PARTITION procedure, on the other hand, always places the pivot value (originally in $A[p]$) into one of the two parititions $A[p..j]$ and $A[j + 1..r]$. Since $p \le j < r$, this split is always nontrivial.

1. Rewrite the QUICKSORT procedure to use HOARE-PARTITION.

Demonstration

At the end of the loop, the variables have the following values:

x = 13
j = 9
i = 10


Correctness

The indices will not walk of the array. At the first check $i < j$, $i = p$ and $j \ge p$ (because $A[p] = x$). If $i = j$, the algorithm will terminate without accessing "invalid" elements. If $i < j$, the next loop will also have indices $i$ and $j$ within the array, (because $i \le r$ and $j \ge p$). Note that if one of the indices gets to the end of the array, then $i$ won't be less or equal to $j$ any more.

As for the return value, it will be at least one less than $j$. At the first iteration, either (1) $A[p]$ is the maximum element and then $i = p$ and $j = p < r$ or (2) it is not and $A[p]$ gets swapped with $A[j]$ where $j \le r$. The loop will not terminate and on the next iteration, $j$ gets decremented (before eventually getting returned). Combining those two cases we get $p \le j < r$.

Finally, it's easy to observe the following invariant:

Before the condition comparing $i$ to $j$, all elements $A[p..i-1] \le x$ and all elements $A[j+1..r] \ge x$.

Initialization. The two repeat blocks establish just this condition.

Maintenance. By exchanging $A[i]$ and $A[j]$ we make the $A[p..i] \le x$ and $A[j..r] \ge x$. Incrementing $i$ and decrementing $j$ maintain this invariant.

Termination. The loop terminates when $i \ge j$. The invariant still holds at termination.

The third bit follows directly from this invariant.

Implementation

There's some C code below.

C code

#include <stdbool.h>

int hoare_partition(int A[], int p, int r) {
int x = A[p],
i = p - 1,
j = r,
tmp;

while(true) {
do { j--; } while (!(A[j] <= x));
do { i++; } while (!(A[i] >= x));

if (i < j) {
tmp = A[i]; A[i] = A[j]; A[j] = tmp;
} else {
return j;
}
}
}

void quicksort(int A[], int p, int r) {
if (p < r - 1) {
int q = hoare_partition(A, p, r);
quicksort(A, p, q + 1);
quicksort(A, q + 1, r);
}
}