[Teammetrics-discuss] Commit Stat.

[Andreas Tille]

Hi,

On Thu, Oct 20, 2011 at 01:13:33PM +0530, Sukhbir Singh wrote:
> Sorry for replying late, I was busy but now, it's the time to work on this!

No Problem.
 
> So there seems to be no way of getting the size of the logs. And even
> if we do get them: the revisions we are having a problem with are not
> binary data but proper revisions. Just to be sure, check this out:
> 
>     svn diff -r 76326 svn://alioth.debian.org/pkg-perl/
> 
> Size: 211M	pkg-perl
> 
> And you will notice that it is not binary data but a 'proper' commit.

Hmmm, I have not checked this out but whatever it is I have *real* doubt
that we should count those > 1 Million lines contributions as something
which is "normal" team work.  Can you look whether there is a certain
amount of commits > 1MB ?  IMHO if we would regard this as some normal
contribution it might perfectly blur what we really want to see in the
stats.  At least I doubt that the committer has written those lines of
code himself in the time frame of this and the previous commit.

> Also, isn't it very uncommon for projects to push binary files to a
> repository? I know we did it but then I doubt that it is possible for
> Debian package repositories, right?

Well, it just happens and I was wild guessing what reasons might have
lead to those huge commits.

> Because we have always aimed that we are not going to exclude
> anything, I am now going to try the file approach I was talking about,
> while setting a buffer size so that the output is flushed to the file
> immediately. I am not 100% sure it will work, but I have hope so it's
> worth trying :) Will keep you posted.

I will not try to stop you from realising your idea.  However, I would
spend some time into checking what *really* was the content of the
commit.  There might be some kind of automated editing (sed s/foo/bar/g)
or so which would mean the real commitment to the team work is rather
one line and not > 1 million of lines.  So your attempt to gain
perfectness should be undergone some reality check whether it really
helps to have a perfectly correct number which does make perfectly no
sense.

Kind regards

      Andreas.

-- 
http://fam-tille.de