Folks, I have just joined this company which has a huge source tree based upon JSP/Servlet and EJB 1.2. No documentation exists. The code has been written over seven years, with a large number of undocumented changes. Are there any tool tah can assist me in tracing the execution? Putting a breakpoint is not helping me much.
+2
A:
Why are breakpoints not helpful? Stepping in the code with the debugger should work. Whether the code is spaghetti or not shouldn't affect the "debugability" of the system.
On how to deal with this mess I suggest writing tons of unit tests for the existing system. It'll allow you to understand the program better and be in a better situation for refactorings as soon as these are needed (obviously very soon). Have a look at http://amzn.com/0131177052
cherouvim
2010-05-19 06:36:31
It could be hard to find units that are testable if it's one large moloch.
tangens
2010-05-19 06:40:08
Definitely! That's where that book helps. In dealing with old large and ugly codebases.
cherouvim
2010-05-19 06:47:36
breakpoints are not helping since most of the method call stacks are very deep. business logic in jsps and javascript add to the difficulty. i was wondering if there could be a stack trace of which method called what (much like looking at log.trace statements if the code were written well), i can just skim through the the sequence to figure out what is the story. Profiler does produce a list of all methods called, but not in any particular sequence.
Amarsh
2010-05-19 06:55:08
I also think that book (Working Effectively with Legacy Code) can be really useful.Some profilers produce a call tree (once I used eclipse TPTP that way) but I doubt it can be of big help.
Andrea Polci
2010-05-19 07:15:22
All IDEs can do that. When you reach a breakpoint there is a "Call Stack" panel where you can see the stack trace leading to this specific line of code.
cherouvim
2010-05-19 07:18:10
@cherouvim I don't think he's talkig about the call stack.What I was meaning talking about TPTP is to produce a tree of all the calls made during an execution, not just the stack of the calls actually waiting to be completed.
Andrea Polci
2010-05-19 08:09:14
Folks, looks like there arent too many off the shelf solutions. I am going the good old way of writing a shell based "code crawler" and figuring out who is calling what. I will post my script if it works for your reference. Great thanks for all your responses
Amarsh
2010-05-20 00:54:17
+2
A:
The good old trick can help here if you're allowed to edit the code: put many System.err.println() at strategic points. It shows the flow of the program, which is probably the first step to discover unknown code.
The trace can also display some variable values or even a stack trace (use new Exception().printStackTrace(System.err)). To avoid a flood of messages, the trace can be guarded by a pre-condition that executes the println only if it worth it.
Be sure to put in each message the current class and method to reference. The message shows clearly the location of the println code, and it will help a lot to remove all the traces when you're done!
kabado
2010-05-19 08:28:56
A:
Great thanks for everyone's inputs. It was a wonderful learning experience. I ended up writing my shell script, which produces a html report. I am attaching the complete file here.
Please note that I am not a regular shell programmer, and I was working on this after hours .. hence the standard of the code is not too good. It has plenty of cut/past jobs off the internet. It works however, and presents the approach you may take to go through your sphegatti code.
Regards Amarsh
#!/usr/bin/bash
# check the number of command line arguments
clear
echo "### CodeCrawler starting"
# test input parameters
if [[ $# < 2 ]]; then
echo "usage: % crawl inputFile/inputDir outputDir"
exit -1
fi
# the working directory is C:\CodeCrawler
cd /cygdrive/c/CodeCrawler
# find all files tha require analysis
if [ -d $1 ]; then
find $1 | grep "\.java$" > allFiles$2
find $1 | grep "\.jsp$" >> allFiles$2
find $1 | grep "\.htm$" >> allFiles$2
find $1 | grep "\.html$" >> allFiles$2
else if [ -f $1 ]; then
find $1 > allFiles$2
fi
fi
# get total no. of files to be scanned
totalFiles=$(cat allFiles$2 | wc -l)
scannedNoOfFiles=0;
echo "### No of files to scan : $totalFiles"
# create the index.html file
rm -rf $2; mkdir $2;cd $2
echo "<html><body bgcolor=\"#ccffff\"><h3>$1</h3>" > dir.html
# crawl through the entire directory
for rootFile in $( cat ../allFiles$2 ); do
scannedNoOfFiles=$((scannedNoOfFiles+1));echo;echo "### Scanning $scannedNoOfFiles / $totalFiles"
# create a dir for the output
rootFileDir=$(echo $rootFile | tr '/' '\n' | tail -1).dir
echo "### Storing output in $rootFileDir"
rm -rf $rootFileDir
mkdir $rootFileDir
cd $rootFileDir
# append to the index.html file
rootFileDirName=$(echo $rootFile | tr '/' '\n' | tail -1)
echo "<a href=\"$rootFileDir/index.html\" target=\"fileFrame\">$rootFileDirName</a><br>" >> ./../dir.html
# obtain all external jsp references
touch jsp.cwl
cat $rootFile | grep "\.jsp" | tr "'\"\?<>=,()[] " '\n' | sed 's/\.\.//g' | grep "\.jsp" | grep -v "http" | sort -u > tmp
for line in $(cat tmp);do
echo /$line | sed 's/\/\//\//g' >> jsp.cwl
done
# obtain all external js references
touch js.cwl
cat $rootFile | sed 's/\.jsp//g' | grep "\.js" | tr "'\"\?<>=,()[] " '\n' | sed 's/\.\.//g' | grep "\.js" | grep -v "http" | sort -u > tmp
for line in $(cat tmp);do
echo /$line | sed 's/\/\//\//g' >> js.cwl
done
# obtain all external css references
touch css.cwl
cat $rootFile | grep "\.css" | tr "'\"\?<>=,()[] " '\n' | sed 's/\.\.//g' | grep "\.css" | grep -v "http" | sort -u > tmp
for line in $(cat tmp);do
echo /$line | sed 's/\/\//\//g' >> css.cwl
done
# obtain all external htm references
touch htm.cwl
cat $rootFile | grep "\.htm" | tr "'\"\?<>=,()[] " '\n' | sed 's/\.\.//g' | grep "\.htm" | grep -v "http" | sort -u > tmp
for line in $(cat tmp);do
echo /$line | sed 's/\/\//\//g' >> htm.cwl
done
# obtain all database references
touch db.cwl
cat $rootFile | grep -i "select.*from" | sed 's/from/\nfrom/g' | sed 's/FROM/\nFROM/g' | grep -i "from" | sed 's/from//g'| sed 's/FROM//g' | awk '{print $1}' | tr '[;"]' ' ' | uniq > db.cwl
cat $rootFile | sed "s/.prepareStatement(\"/\nX_X_X/g" | grep "X_X_X" | sed "s/X_X_X//g" | tr '[ ,\$ ]' '\n' | head -1 | uniq >> db.cwl
# obtain all references to java classes. we include everything with signature com. and exclude "www" and "flight"
cat $rootFile | tr '["=%;/<>@\t) ]' '\n' | grep "com\." | grep -v "codepassion\." | grep -v "www" | grep -v "flight" | sort -u > tmp
echo > tmpDirectReferences
cat tmp | grep "(" >> tmpDirectReferences # directReferences are like au.com.mycompany.servlet.MiscServlet.getCckey()
echo > tmpDirectReferences
cat tmp | grep -v "(" >> tmpJavaFiles # javaFiles are like Person aPerson; ... aPerson.getPolicy()
# read directReferences and produce the class.cwl file by identifying class and method
echo "#D# Looking for direct references"
while read classLine; do
methodName=$(echo $classLine | tr '\.' '\n' | tail -1 | sed 's/(//g')
className=$(echo $classLine | sed "s/\.$methodName(//g" | tr '[()]' ' ')
echo $methodName >> $className.cwl
echo "### class: $className method:$methodName"
echo $className >> tmpDirectReferencesReformed
done < tmpDirectReferences
# read javaFiles every fully qualified class name and grab the class from it. then grab the method from it
echo "#J# Looking for indirect references"
while read classLine; do
className=$(echo $classLine | tr '\.' '\n' | tail -1)
echo "#F# find: $classLine"
# indirect references are in the form className objectName ... and then objectName.methodName
cat $rootFile | grep "$className .*;" | sed -e "s/$className[ \t]\+\([a-zA-Z0-9_]\+\)[ \t]*[;=].*/\1/g" | sed 's/^[ \t]*//;s/[ \t]*$//' | sort -u > tmp$ClassName
# read tmp$className and find all properties and method references
while read methodLine; do
cat $rootFile | grep "$methodLine\." | tr '[ (]' '\n' | sed "s/$methodLine\./\n$methodLine\./g" | grep "$methodLine\." | sort -u | grep -v "[\"%]" | grep -v ".com." | tr '.' '\n' | grep -v "$methodLine" >> $classLine.cwl
done < tmp$ClassName
# direct references are className.methodName
cat $rootFile | grep "[ ()\"']$className\." | tr ' (' '\n' | grep "$className" | tr '.' '\n' | grep -v "$className" >> $classLine.cwl
cat $rootFile | grep "$className\." | tr ' (' '\n' | grep "$className" | tr '.' '\n' | grep -v "$className" >> $classLine.cwl
done < tmpJavaFiles
# consolidate all information to generate the html files
echo "### Generating index.html"
rootFileName=$(echo $rootFile | tr '/' '\n' | tail -1)
touch index.html
echo "<html><head><title>$rootFileName</title></head><body bgcolor=\"#ffffcc\">" >> index.html
echo "<h3>$rootFile</h3>" >> index.html
# put all java classes
echo "<br><h3>Referenced classes</h3>">> index.html
cat tmpDirectReferencesReformed | uniq >> tmpJavaFiles;cat tmpJavaFiles | uniq > tmpJavaFilesU; mv tmpJavaFilesU tmpJavaFiles
while read aLine; do
echo "- <a href=\"$aLine.html\" target=\"methodFrame\">$aLine</a><br>" >> index.html
done < tmpJavaFiles
# put all DBs
echo "<br><h3>Referenced Tables</h3>">> index.html
while read aLine; do
echo "- $aLine<br>" >> index.html
done < db.cwl
# put all JSPs
echo "<br><h3>Referenced JSPs</h3>">> index.html
while read aLine; do
echo "- $aLine<br>" >> index.html
done < jsp.cwl
# put all JSs
echo "<br><h3>Referenced JavaScript</h3>">> index.html
while read aLine; do
echo "- $aLine<br>" >> index.html
done < js.cwl
# put all htms
echo "<br><h3>Referenced htm</h3>">> index.html
while read aLine; do
echo "- $aLine<br>" >> index.html
done < htm.cwl
# put all css
echo "<br><h3>Referenced css</h3>">> index.html
while read aLine; do
echo "- $aLine<br>" >> index.html
done < css.cwl
echo "</body></html>" >> index.html
# generate a html for each class file and put all accessed methods in it
for aLine in $( ls *.cwl ); do
cat $aLine | uniq > tmp; mv tmp $aLine
fileName=$(echo $aLine | sed 's/\.cwl//g')
echo "#G# generating $fileName.html"
echo "<html><body bgcolor=\"#ffddee\">" >> $fileName.html
echo "<h3>$fileName</h3>" >> $fileName.html
for bLine in $( cat $aLine | sort ); do
echo "$bLine<br>" >> $fileName.html
done
echo "</body></html>" >> $fileName.html
done
# cleanup and return
#rm *.cwl *tmp*
cd ..
done
echo "</body></html>" >> ./dir.html
rm ../allFiles$2
echo "### CodeCrawler finished"
Amarsh
2010-05-28 00:37:36