next up previous contents index Search
Next: Analysis Up: 0.3.1 The Quicksort Previous: Picking a Pivot Partitioning

Now, here is code to do the actual work of dividing an input range into two partitions based on a pivot point. Again, it is written in C:

int partition (data *array, int left, int right, int pivotval) {
  do {
    swap (&array[left], &array[right]);
    while (value(array[left]) < pivotval) left++;
    while (value(array[right]) >= pivotval) right++;
  } while (right >= left);

  /* this will be the value of the first element in the right part */
  return (left);

The above routine cleverly moves inward from each end of the input range swapping data values that are on the wrong side of the pivot value until the two inward-moving indices meet in the middle. The initial swap, above, is not necessary; it is included only to make the do-loop more simple. In the worst case, n-1 comparisons and $\frac{n}{2}$ swaps are necessary to partition the data set.

Scott Gasch