#43 Exclude specific commits or users
(I'll file this as a place for ideas and discussion)
A common problem is that the LOC per author charts get
messed up because someone moved or renamed lots of
files, merged another branch or added lots of code from
somewhere else, thus getting unjustified credits for many
lines of code. In such a case, the LOC per author chart is
effectively rendered useless.
A possible fix: Provide a way to exclude specific commits
from all comparative author charts and tables. You could
exclude any "bad" commits. The charts would be nice and
competitive again.
But I have some qualms. This smells a bit like "I'll fix the
numbers until the charts prove my point". Providing a way
to easily exclude inconvenient data would be bad for the
integrity of StatCvs. This is a statistical tool, not <blow
altitude="low">PowerPoint</blow>.
Logged In: NO
True, but there are specific commits that *SHOULD NOT* be
counted towards the author's LOC count. For example, the top
two developers by LOC on our CVS
(http://www.cs.trinity.edu/~jmartin5/cvs/) both committed
.jar files containing either Graphics or Unit Testing
Libraries. This skews the results unneccesarily.
Jess Martin
jess.martin@trinity.edu
Logged In: YES
user_id=584620
Jess, you should always check in binary files with the -kb
option of CVS, otherwise they can get messed up by the
keyword substitution feature of CVS. Also, if the files have
been correctly checked in as binaries, StatCvs will ignore
them when counting lines. Furthermore, in the next version of
StatCvs, you will be able to exclude specific file types by
using a switch like '-exclude **/*.jar' on the command line.
Logged In: NO
Man, I also have this problem :( We use colaborative open
source software to learn and let things clean and stimulate
programming but sometimes it can be a hassle :( People here
uses tons of separate code, then we need to CVS IMPORT them.
When we import tons of lines, you can expect the result...
It would be VERY important, for all communities to use, if
there is support for user exclusion (like statcvs
--exclude-users cvs,admin). For now, we are hiding certain
start dates (we import all of them once at day 1, for
example, and do a global statcvs FROM day 2, and weekly
statcvs graph generation of day 2-8, 9-15, and so on...
Logged In: NO
Comment above made by ivanrolim - ivanrolim@yahoo.com.br
Logged In: YES
user_id=584620
Originator: YES
Previous poster filed a duplicate with more rationale at http://sourceforge.net/tracker/index.php?func=detail&aid=1557187&group_id=57558&atid=484572
Logged In: YES
user_id=584620
Originator: YES
Excluding users is now possible with by adding "-no-developer foo".