This is, indeed, very cool!
Some points to ponder for your thesis:
(a) While the numbers seem to agree with intuition, the development of "review indicators" and their weightings will provide a way of "explaining" the aggregate number (in terms of its underlying indicators). This will help people understand/believe the system.
(b) The modules that you can't provide a number for are also quite interesting! Think about what that implies for your system--is it a deep architectural limitation, or simply the lack of an appropriate review indicator? Now, consider the case of hackyDocBook, which has no Java code but a lot of XML, and say you develop a review indicator for XML. Is it fair to rank hackyDocBook along with, say, HackyKernel? Why or why not?
(c) It might be interesting to interview Hackystat developers about code and get their personal opinions about where things are ranked, and _why_ a certain module should or should not be reviewed. This might provide additional perspectives on potential review indicators.
Cheers, Philip
--On Thursday, March 24, 2005 3:13 AM -1000 Aaron Kagawa <[EMAIL PROTECTED]> wrote:
Hey Guys,
You all seemed a little interested in the PRI rankings for the top level modules. So here is a complete sorted list. The higher the number the better. There are a few ? (missing) rankings. Some of these modules are ignored, some of them have no FileMetric data so the system can't tell if they are no longer existent, some of them have no java files, and some are just plain missing for some reason.
Anyway.. You'll see that there is an interesting ordering right off the bat.. hackyStatistics, hackyReport, and hackyKernel grouped together. In my opinion this seems almost perfect. hackyStatistics is probably our most stable package, followed by hackyReport, then followed by hackyKernel. Also, if it wasn't for Mike's good programming, hackyStdExt would be right next to that group. The cool thing is that you can almost see the kernel-lized architecture we have.
In hackyCGQM defense.. it is really brand new code. and it really probably should be reviewed. So, its ranking seems appropriate.
thanks, aaron
hackyStatistics 933.00 hackyReport 872.25 hackyKernel 870.71 hackyHPCS 865.50 hackyStdExt 827.21 hackyVCS 804.75 hackyReportExample 792.00 hackyDependency 785.25 hackyAnt 781.73 hackyPerf 773.75 hackyTDD 758.25 hackyTelemetry 748.08 hackyVim 688.00 hackyReview 685.18 hackyIssue 602.00 hackyCli 587.60 hackyJBuilder 552.00 hackyEclipse 533.67 hackyJupiter 500.00 hackyCGQM 486.68 hackyVisualStudio ? hackyTelemetryViewer ? hackyVIM ? hackyDocBook ? hackyCourse ? hackySQI ? hackyPrjSize ? hackyBuild ? hackyMDS ? hackyEmacs ? hackyJPLBuild ? hackyJira ? hackyCLI ? hackyPRI ? hackyLoadTest ? hackyEstimate ? hackyCocomo ? hackyOffice ?
