Algorithms and OOD (CSC 207 2014S) : Labs

Laboratory: Debugging with Eclipse


Summary: We begin to explore the ways in which we can use a debugger to better understand flaws in our code.

Preparation

In the the laboratory on unit testing, you forked and cloned the repository https://github.com/Grinnell-CSC207/lab-unit-testing. You'll work with that same repository. (So return to the directory if you have it, and make a new copy if you don't.)

Exercises

Exercise 1: Removing A's

As you may have noted in the the laboratory on unit testing, the procedure SampleMethods.removeAs is not quite successful in its attempt to remove all copies of the letter 'a' from its parameter string.

If you haven't yet written your test cases, here's one.

public void testRemoveAs() 
{
  assertEquals("empty string", "", "");
  assertEquals("no as", "hello", "hello");
  assertEquals("eliminate one a", "", SampleMethods.removeAs("a"));
  assertEquals("eliminate many as", "", SampleMethods.removeAs("aaaa"));
  assertEquals("eliminate one a, short string", "pin", 
               SampleMethods.removeAs("pain"));
  assertEquals("eliminate many as, medium string", "lphbet", 
               SampleMethods.removeAs("alphabet"));
  assertEquals("eliminate many as, silly string", "BCDEFGHIJKLMNOPQ",
               SampleMethods.removeAs("aBaaCDaaaEFGaaaaHIJKaaaaLMNaaaOPaaQa"));
  assertEquals("eliminate prefix and suffix as", "bbb",
               SampleMethods.removeAs("aaabbbaaa"));
} // testRemoveAs

You may be able to tell by inspection why the method fails. But let's assume that you don't.

Open the code for removeAs and right click in the grey bar to the left of the code to set a breakpoint at the start of the method.

Create a new main class that duplicates the failed calls to removeAs from your unit tests. Select Run > Debug As > Java Application.

A dialog box should pop up asking you to confirm switching to the Java perspective.

If all goes well, Eclipse should stop at the point that you inserted a breakpoint.

a. What do you expect to happen if you click the Resume button - the button that looks like a green triangle. (Note that in the future, you can also hit F8.)

b. Check your answer experimentally.

As you may have noted, Eclipse resumed computation and ran until the completion of this program. (Presumably, with incorrect output.) To see the results, you may need to switch back to the Java perspective. You can get that perspective by clicking on the downward arrow in the upper-right-corner of the screen.

c. Start the program again. This time, let's single step through the procedure, using the Step Over (also F6). See if you can identify where the code goes wrong.

d. Correct the code to the best of your ability, remove the breakpoint, run the unit tests again, and see if your code passes all of the tests. If so, go on to the next exercise. If not, repeat the debugging steps until you find the next bug.

Exercise 2: Removing B's

The removeBs procedure has much the same goals as removeAs although it uses a different (but still buggy) approach.

Use JUnit and the Eclipse debugger to identify and correct the errors.

Note: Your goal is to correct the errors in this approach. Inserting slightly modified code from removeAs is not an acceptable strategy.

Exercise 3: Exponentiation

The SampleMethods.expt method computes xp using a divide-and-conquer approach.

  • x0 = 1
  • x2k = xk * xk
  • x2k = (x2)k
  • xk+1 = x * xk

Some people combine the last two when dealing with an odd exponent.

This approach requires only log2p multiplications to raise x to the pth power, while the naive loop requires p multiplications. (Of course, if you have a book of tables, or functions that simulate those tables, you can compute xp in two table lookups.)

It's a nice approach, but have we implemented it correctly?

If you haven't done so already write unit tests for SampleMethods.expt(int,int).

a. Add the following assertion at the start of your test.

assertEquals("1K", 1024, expt(2, 10));

b. Run the test. It will likely fail.

c. Since the test failed, it will be useful to write a short experiment to just do that one call.

/**
 * A quick experiment with the expt method.
 */
import java.io.PrintWriter;

public class ExptExpt
{
  static final int base = 2;
  static final int expt = 10;

  public static void main(String[] args)
      throws Exception
  {
    PrintWriter pen = new PrintWriter(System.out, true);
    pen.println(base + "^" + expt + " = " + SampleMethods.expt(base, expt));
    pen.close();
  } // main(String[])
} // class ExptExpt

c. Set a breakpoint at the start of the expt method. (Make sure that you choose the right one. There are two!)

d. Start the debugger. It should bring you to the first line of expt.

e. What do you expect to happen if you click the Resume button? (The button that looks like a green triangle.)

f. You may have discovered that instead of returning to the call in the unit test, the debugger continued executing the code until the next call to expt, which is a recursive call. Hit the Resume button another time.

g. You are now three levels deep in the recursive call stack for expt. In the Debug pane, navigate between them to see the changing values of x and p.

h. Single step through the code to see if you can identify where the error occurs.

i. Since intermediate values are not clearly represented in the code, you may find it difficult, if not impossible, to quickly identify the error. So what next? You could explicitly insert temporary values for the recursive call. Instead of calling return in each case, you could set a local values (e.g., results and then exit in the logical case). Or you could get Eclipse to behave better.

Choose one approach and see if you can identify the error. Get help if you're not sure which approach you should use or if you still can't identify the error after trying additional approaches.

For Those With Extra Time

Extra 1: Exponentiation, Revisited

Consider the expt(double, int) method. As you might have noted, it doesn't work any more correctly than the old version of expt

One issue we may hit in unit testing is that doubles are approximate. Hence, slightly different orders of computation can make slight differences in the result (e.g., in practice Math.sqrt(2)*Math.sqrt(2) is often not the same as Math.sqrt(2*2), even though they are logically the same.

Write appropriate unit tests for this alternate version. Then determine if your corrections from the exercise above suffice. If not, use the debugger to figure out why.

Copyright (c) 2013-14 Samuel A. Rebelsky.

Creative Commons License

This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/ or send a letter to Creative Commons, 543 Howard Street, 5th Floor, San Francisco, California, 94105, USA.