Algorithms and OOD (CSC 207 2014F) : Labs

Laboratory: Binary Search in Java

Summary: In today's laboratory, you will explore issues pertaining to search in Java. Along the way, you will not only consider the binary search algorithm, but explore some program design issues in Java.


a. You are likely to find it useful to have the corresponding reading open in another window.

b. Create a new Eclipse project and Java package for this lab. (I'd recommend that you also create a Git repository, but it's up to you.)

c. Create a new class, Utils, that will hold much of code that you will write today.


Exercise 1: Binary Search in Arrays of Integers

Although the reading introduced a variety of techniques for designing generalized search algorithms, it's probably easiest to start by focusing on a single type.

Implement the following procedure.

 * Search for val in values, return the index of an instance of val.
 * @param val
 *   An integer we're searching for
 * @param values
 *   A sorted array of integers
 * @result
 *   index, an integer
 * @throws Exception
 *   If there is no i s.t. values[i] == val
 * @pre
 *   values is sorted in increasing order.  That is, values[i] <
 *   values[i+1] for all reasonable i.
 * @post
 *   values[index] == val
public static int binarySearch (int i, int[] vals) 
  throws Exception 
  return 0;   // STUB
} // binarySearch

Exercise 2: Testing Our Algorithm

Evidence suggests that (a) many programmers have difficulty implementing binary search correctly and (b) many programmers do only casual testing of their binary search algorithm. But it's really easy to write a relatively comprehensive test suit for binary search.


Implement this test. Then repair any bugs you find in your implementation of binary search.

Note that I've found this test very useful. A surprising number of pieces of code fail just one or two of the many assertions in this test.

Citation: This test is closely based on one suggested by Jon Bentley in a Programming Pearls column.

Exercise 3: Care In Checking Midpoints

As binary search is phrased in the reading, when we note that the middle element is not equal to the target value, we either set ub to mid-1 or lb to mid+1. But programmers often get confused by the need for the +1 and -1.

Determine experimentally what happens if you leave out the +1 and -1. Explain why that result happens.

Exercise 4: An Alternate Approach

In implementing binary search, you either wrote a loop or a recursive procedure. Write a second version of binary search that uses the other approach.

Exercise 5: “Timing” Search

In theory, binary search should take O(log2n) steps. Does it really? Augment each of your methods so that it counts the number of repetitions (loop) or calls (procedure). It's probably easiest to create global variables that you set to 0, and then increment at the top of the loop body or at the start of the procedure.

Build some moderately large arrays (at least 1000 elements) to verify that you get the expected running times.

Exercise 6: Searching for the Smallest Value

a. Implement the following procedure:

 * Find the "smallest" integer in an array of integers
public static Integer smallest(Integer[] values, Comparator<Integer> compare) 
   return null; // STUB
} // smallest(Integer[])

b. Run your procedure with a comparator that does the standard integer comparison.

c. Run your procedure with a comparator that does reverse integer comparison (e.g., if x < y, compareTo(x,y) should return a positive number.

d. Run your procedure with a comparator that does closest-to-zero comparisons.

For Those With Extra Time

Implement a generic binary search that takes a comparator as a parameter.

public static <T> int binarySearch(T value, T[] values, Comparator<T> compare) 
  throws Exception 
} // binarySearch