12.WebSpiderReview

The goal of reviewing a classmate's WebSpider assignment is to improve the ability to read, write, and test Java systems.

1. Installation Review
This process should be fairly smooth since we all used the same IDE and the same tools to write and test our code. I downloaded my classmate's software and extracted it into my Eclipse workspace. The packages contains two source code files (WebSpiderExample and TestWebSpider), a lib directory with checkstyle, emma, findbugs, and pmd sub directories, and various build and documentation files. The package had all the necessary files. Therefore, I was able to import his project into Eclipse without any effort.

The command ant -f verify.build.xml returned no errors and completed successfully. Because, I was a little bit suspicious, I decided to execute each build file individually but no errors were found. However, emma block coverage is only 83% and the line coverage is only at 85%.

I was able to build a jar file and execute his code with java -jar webspider-ivanw.jar -totallinks http://www.hackystat.org 100. As read from the console, it seemed that some Websites refused the connection. However, this did not bother the code execution and there were 1585 links found. At this time I can not be certain if this number is true since my code returns a slightly lower number of links (Read on).

2. Code format and conventions review
After reviewing the code, I recommend the following formatting corrections:

FILE	LINES	VIOLATION	COMMENTS
WebSpiderExample.java	46	EJS 6	Spelling ("of" twice)
WebSpiderExample.java	58, 161	EJS 33	Keep comments and code in sync
WebSpiderExample.java	71, 174, 183, *	EJS 7	use space between code blocks
WebSpiderExample.java	150	EJS 49	Omit the subject in summary descriptions of actions or services
TestWebSpiderExample.java	3	ICS-SE-Java-2	Do not use wild card "*" in import statements
TestWebSpiderExample.java	22	EJS 33	Keep comments and code in sync

3. Test case review

Black box

Ivan's TestWebSpiderExample.java tests all methods within WebSpiderExample through calling the main method. Since all methods are in the same class, it seem like the test cases cover all methods written. Form a black box perspective this might be enough. But I would suggest covering all if-then-else statements as well.

In TestWebSpiderExample.java there are 4 calls to the main method each testing with different number of parameters but the same values. In addition there is one call to testMainMethod.

A nice extension to the tests that he already has would be testing for -totallinks and -mostpopular with 0 pages. Then testing it with a high number of pages and later with some illegal value such as, providing string input instead of number. He also could test with an empty parameter. For example, the URL String, or any other required parameter, could be empty/null. In addition a self made website could be set up to see if the code returns the correct values.

Since only the one call to testMainMethod has a assert statement I would suggest putting more assert statements and maybe calling testMainMethod on all tests since this method returns a value that can be compared for correctness.

White box

The emma coverage report returned 83% of block and 85% of line coverage, which does not seem satisfying from a white box perspective.

Analyzing emma's coverage report, I would suggest to include testing for the following:

1 - The length of linksVisited array is set to 65535. Since all test cases test only 5 pages, the linksVisited will never reach over 65535 links. Therefore, a test case must be set up or the length should not be preset.

2 - Various catch boxes are not covered. My suggestion is to set up a Website that would satisfy covering all the exceptions.

3 - Set up a test case were the user enters something else than the supposed -totallinks or -mostpopular option, and one of the else statement would have sufficient coverage under emma.

Break da buggah

I have set up my own very basic test website. The expected test results for -totallinks should be 3 and -mostpopular should return http://www2.hawaii.edu/~michalz/WebSpiderTestWebSite.html with 2 links when the maximum number of pages to traverse is set to 3 or higher.

Running Ivan's code on it returned incorrect results.

command: java -jar webspider-ivanw.jar -totallinks http://www2.hawaii.edu/~michalz/WebSpiderTestWebSite.html 3

output: 5

Changing the number of pages to traverse from 3 to 5 returned 9.

command : java -jar webspider-ivanw.jar -mostpopular http://www2.hawaii.edu/~michalz/WebSpiderTestWebSite.html 3

output: The most popular page is: http://www2.hawaii.edu/~michalz/WebSpiderTestWebSite.html with 0 pages that point to it.

comment: Right website, wrong number of pages.

Setting the maximum pages to traverse to a different number, gives all kinds of results with one of them being the right one.

command: java -jar webspider-ivanw.jar -totallinks http://www.myspace.com/shakadownstreet 10

output: crash - Sys.Application.notifyScriptLoader(); 'failed: SyntaxError: illegal character

4. Summary and lessons learned
I have learned to think and design better test cases for programs. Most people, including me, tried to get emma coverages as high as possible without even thinking if the tests we write are actually helping to test the actual program execution. I have also learned that it is important to set up an own testing environment before testing the code for correct output. For example, if I would set up a test website for WebSpider right from the beginning, I could be more certain that my code does what it supposed to be doing. I have also learned that using someone's test website isn't very good since I found one link on Randy's website that was down, and therefore my code returned a different answer than expected. If you know what your code does, than it is easy to set up a testing environment and write more sufficient test cases.

Extra Credit (revised)

Source Code: Available here

So, I did some research on the httpunit package and I found a nice article named "introduction to Httpunit". From, this article I have learned that we must set the script error exception to false. However, this did not solve my problem crawling the myspace Website. Then I found other useful Website that helped me learn and understand the Httpunit package. However, most of the other articles did not help in solving the problem. Since you are allowed to set options in Httpunit as in : HttpUnitOptions.setExceptionsThrownOnScriptError(false);
I then looked, with help of the Javadoc in Eclipse, at setting other options. When I set HttpUnitOptions.setScriptingEnabled(false);
the problem of crawling the myspace site was solved.

11 WebSpider (revised)

Source Code: Available here

First of all, it was a good experience having this assignment extended because I was able to improve my logging feature which is implemented with java.util.Logging.Logger. Although, my web spider was working already before, I decided to revise the code and add more security to it. Therefore, I have surrounded necessary code with try and catch blocks. The only problem associated with it is when running Emma on it. Since the code won't always go through the catch block, Emma will complain about it and the coverage drops down from the expected 100%. But at least it won't stop the code execution when a link is down or when a website is not accessible. In addition, I think that writing Junit tests isn't always very easy. Sometimes it seems like Junit tests are not necessary for certain blocks of code but then Emma coverages will drop. So, to satisfy this I must write a test which just goes through a block which only contains a print statement. I have also found a way of turning off the scripting errors in httpunit and therefore my code works for nearly all web pages, including myspace.com.

Extra Credit

I have tried to accomplish the extra credit assignment by playing with the code until it would actually execute the myspace site without any exceptions. However it always complained about ASP.NET AJAX Framework is missing. While trying to make it work I found that httpunit has a function called HttpUnitOptions.setExceptionsThrownOnScriptError(false); that would skip all the scripting but this did not help either. The function takes either true or false as the parameter whether you want scripting errors to be printed or not. Then I went on to Google to get some more information on how to solve the problem however in the end I was not able to execute myspace without any errors. At leat my web spider works for a lot of other site.

11. WebSpider

Source Code: Please click here to download

I was able to accomplish all the tasks. However the third task which required us to implement logging, I was able to completed it only partially since I choose to put a bunch of System.out.print line statements. I could have implemented a better logging, but I did not have the time to finish it with a complete logging application before the due date. Therefore, I realized that it is important to start on harder assignments at least a week before the due date.

Write the web crawler was for me the hardest part. Since the httpunit package doesn’t come with all the Javadoc comments which are always very useful when writing Java code in Eclipse. I also missed some helpful documentation or examples for httpunit on Google or even on their own Website. If I would be more familiar with the package then I am sure that I could write better and more robust code then the one which is available here. That is what I think that writing the actual web crawler made more difficult.

A problem that I ran into in the very beginning was:

When I issued the command ant run, it gave me an error saying that Junit.org does not exist. It took me a while to figure out that putting Junit into the fileset in the build.xml file would actually solve my problem. After that it ran fine, but I ran into problem with the httpunit package.

Finally, I was able to get a good coverage from emma which was near 100% for all. I was not able to get the full 100% because PMD complained about ArrayList which I could not solve in the short time.

In conclusion, I think that this assignment was helpful in dealing with unfamiliar packages and it also gave me a chance to brush up on recursion. Once I heard from a senior worker that he has never used recursive methods in his career when writing programs. I would agree that recursion is not necessary since it is heavier on the memory, has slower execution (as can be observed when executing this code) and is more prone to errors.

By the way, be patient when executing this code since it may take a while until it recursively traverses through the Web links until a specified number of pages have been visited.

10. Stack

Source Code: Please click here to download

This assignment was not as easy as it seems when reading the assignment instructions. Various tasks had to be mastered in order to complete this assignment including learning the Ant build system as well as several Java automated quality assurance tools. Personally, I was glad when at the end all the specified tasks were successfully working.

I have encountered a few major problems while I was working myself down the instruction list.

problem - 1
I was able to download and install SCLC Code Size Counter correctly, however when I tried to invoke it via command line it gave me an error. It turned out that I did not have Perl 5 installed on my machine.

solution - 1
I downloaded ActivePerl and the problem was resolved. At this point I was able to execute all the targets in Task 1 but verify, without errors.

problem - 2
Another problem that I ran into was associated with javaNCSS. The download and installation were rather easy but I found the actual instruction on kclee.de a little difficult to follow. However, the actual problem was that I had to learn how to write the build.xml file for javancss.build.xml, which was a little more challenging than all the previous tasks since I did not write any XML before. Fortunately, I know HTML which did help a little.

solution - 2
In order to complete the required tasks for the javancss.build.xml file, I looked at and studied all the other build.xml files that are included in stack-5.0.909.zip. In addition, I used Google to search for related topics.

ANT vs. MAKE

I remember writing several MAKE files and having the "tab" problem because it is necessary for MAKE to have proper indentation. When working with ANT I did not have this kind of problem. However, in order to write the necessary files for ANT tasks I was forced to learn XML. After all, I believe that ANT is a more powerful tool and when allowed I would chose it over MAKE. On the other hand, knowing a scripting language such as XML is very important for the future.

SCLC vs. JavaNCSS

Both, SCLC and JavaNCSS can be used out of an ANT build.xml file. Currently, SCLC supports 27 different file types while JavaNCSS is designed specifically for the Java programming language. (I believe) Therefore, JavaNCSS looks more at packages, objects, and functions while SCLC looks at files with various extensions and concentrates on lines, comments, etc. JavaNCSS provides the number of javadocs generated whereas SCLC gives an overview of the number of files with different file extensions, and much more. Therefore, I believe that both are good tools but most powerful when they are used together on a project.

Conclusion

This assignment, as many other ICS-projects, was time consuming. Therefore, it is important to allocate enough time to be able to learn all the useful tools. At first, this assignment may seem a little difficult, however after mastering all the basic readings on all the tools involved, it is rather understanding than confusing. Finally, it would safe me a lot of trouble in my previous programming courses, when I would be aware of these open source Java automated quality assurance tools.

08. CodeRulerRedux

Group members: Michal Zielinski & Chiao-Fen Zielinski-Liu

Source Code: Please click here to download

This time we had more time to think about a better strategy for our CodeRuler project. After reading other classmates' code we found a better way to claim land. Jared has an interesting loop in his CodeRuler code that loops through the whole grid starting in the left upper corner. We implemented a similar strategy. However, we divide the grid into four equal squares and assign peasants to each square. Since the fourth square gets less peasants and therefore they cannot finish claiming all the land, once the peasants in the first square finish claiming the land we assign them to help peasants in the fourth square. In the meantime, our knights are instructed to capture all castles that we don't own and anything that is in their way. After that the knights will capture all enemy peasants. When all enemy peasants are captured the knights go on to capture the enemy knights. We instruct our knights to capture all the castles first in order to be able to create more knights when needed in the battle with other knights. Therefore, peasants are only created when we have twice as many knights as other knights or when we don't have any peasants. This strategy works out pretty good against all the sample rulers. We were also able to include Java Documentation and Coding Standards in our CodeRuler source code which makes our code more readable and understandable to another reader.

This are the result runs:

	Result 1	Result 2	Result 3
michalz-chiaofen vs. migrate ruler	854 - 0	827 - 1	833 - 0
michalz-chiaofen vs. gang up ruler	782 - 184	784 - 140	553 - 399
michalz-chiaofen vs. split up ruler	725 - 256	733 - 230	748 - 182

Writing this code gave me a chance to become more familiar with the Java Documentation and Coding Standards which I learned by reading "The Elements of Java Style". I must say that it is not very easy to follow exactly what the book says since it doesn't have enough examples of certain rules. Hopefully, as this class goes along, I will be able to write code with perfect documentations.