(Latest Revision: 02/24/2008)


Comparing Sorting Algorithms




Write a program that REPEATEDLY:
  1. prompts the user to enter a filename,
  2. opens the file,
  3. reads a comment from the first line,
  4. reads an integer N from the second line (e.g. N = 1000),
  5. reads N integers from the rest of the file, putting them into consecutive locations in an array,
  6. writes the name of the input file,
  7. writes the comment from the input file,
  8. sorts an identical copy of the array using each of the following methods:
  9. writes a report of the number of compares and moves done in each sort, and
  10. asks the user if s/he wishes to continue
UNTIL the user wishes NOT to continue.


MOTIVATION:

Working on this program and actually experiencing the differences in the amount of work and the amount of runtime is a nice way to get familiar with analysis of algorithms. This is an important part of the "science" in computer science.


DETAILS:

Your program must make a little announcement (examples to follow) just before starting and just after completing each run of one of the four sorting functions. This will allow you to get a sense of how long each individual algorithm takes to do its work.

Your program must pass in a new copy of the original array each time it calls a sorting function. We don't want the first function call to be the only one that really sorts the original array!

When you count "moves" count the number of times that an element of the list is copied from one location to another (with an assignment statement). (The book's C++ swap function performs 3 moves.

When counting compares, count only comparisons of list elements (the things that are being sorted), not the array indices. For example, if the array you are using is named "data" the line of code:

if (data[j] < data[j - 1])

contains a comparison that you would have to count, because it is a comparison of two elements in the list. On the other hand, the line of code:

while (i < j)

does not compare elements of the list. It is a comparison of array indices only. Therefore you must not count that comparison.

The line of code

for (;(loc > 0) && (data[loc-1] > nextItem)

contains one comparison that you should not count and one comparison that you should count.

You should not count "loc > 0" because it compares indices. You should count "data[loc-1] > nextItem" because it compares a list element with (a copy of) another list element.

Keep in mind that it may be a little "tricky" to insert the code that counts the comparisons executed in if-clauses and loop conditions. Give the problem due thought and consideration.

There is C++ code for performing sorts here:

ftp://www.cs.csustan.edu/pub/john/2008Src/DAPSWM_C++_ed5/ByName/c09

and Java code here:

ftp://www.cs.csustan.edu/pub/john/2008Src/DAPSWM_Java_ed2/Chapter10/SortsClass.java

You will need to modify this code so that it takes care of the counting of the compares and the moves. That will mean adding parameters to functions (methods) and adding pieces of code that do counting within functions. We can discuss more details of this in class. In the assignment directory, I also included a little prototype of a driver program called driver.cpp You can get ideas from driver.cpp for formulating the program you do for this assignment.


SAMPLE INPUT AND OUTPUT:

Let's look at two sample input files and the corresponding output in order to better understand what the format of the files and outputs has to be. The sample input files are ord50 and ran1000 The corresponding sample program output: sample.out Note: When you see a message such as "selection sort starting ... done" understand that the program first writes "selection sort starting ... " then performs the selection sort, and then writes "done."


TESTING:

Lists can be small, medium, and large in size. They can also be random, ordered, or reverse-ordered. You need to test all the possible combinations (there are 9). When you test your program on large lists it helps you appreciate the efficiency advantages of the advanced sorts. Computers are getting faster and faster! I felt I had to choose lists of size 30,000 for your large sets. A few years ago lists of size 5,000 probably would have been quite large enough. It is very important that you test all the data sets I have indicated. It will determine what you learn from this assignment (and will also determine a large fraction of your grade). I put a package containing all the inputs you need here (but don't click until you are ready to download):

ftp://www.cs.csustan.edu/pub/john/dataFiles.tar.gz

(Some browsers may not access the link above properly. If one browser has a problem, try another. Clicking on the link should download and unpack the files on most systems. You may have to do an extra step to unpack the file.)

In the header of your program, include a paragraph or two describing what you learned (about the performance of the different algorithms) from writing and testing this program. Do your very best to make this well-written. This will be another aspect of the assignment that will count heavily.

While making the test script for this program do not cat the input files onto the screen! This would be the normal procedure if the input files were of manageable size, but most of the input files we are using in this assignment are way too long! If you write the program according to the specifications I gave you, then the information printed by the program will tell me what I need to know.


WHAT TO TURN IN:

Since this assignment does not require you to do a large amount of program design work, I am not requiring you to turn in a preliminary version.

You will turn in two printer outputs (hardcopies) and you will send me one e-mail message. Please follow these rules: Here is the list of things you have to turn in: Understand that I'm asking you to send me all the files I need to compile the program (plus some other things).

Make sure the script you send is cleaned up using the sort of "col -b" filtering trick illustrated here.

Make sure all your source files are clean too. Sometimes files transferred from a PC to a Sun Ultra need to be cleaned up.

Don't send me any compiled code!

E-mail me the shell archive file with the subject line:

CS3100Prog2

Note that there are no spaces in the subject lines given. It is important that you do not insert any spaces. My e-mail address is: john@ishi.csustan.edu.


DUE DATES:

For the due dates, see the class schedule.


QUESTIONS:

Please bring up in class any questions you still have about this assignment.