[Teammetrics-discuss] What is left.

[Andreas Tille]

On Wed, Aug 17, 2011 at 03:25:46PM +0530, Sukhbir Singh wrote:
> > This would be my highest preference.  Please try to adapt David
> > Bremner's code to make sure we really track the commits in debian/.  If
> > anything remains unclear please discuss this on the list (perhaps David
> > in CC).
> 
> Yup! Actually I ran into some problems with the branches/ tags and
> merges that were causing misleading metrics so I have decided to check
> everything thoroughly.

OK.  Just talk about your problems in case they are hard to solve.
 
> > Well, my prefered way is definitely the mbox approach because this is
> > the only thing which makes really sense.  However, I do not expect it
> > ready in five days.
> 
> Well, our converter is ready and so is liststat.py, so this should not
> be much of a problem.

You name it: The word "should" IS the problem.  Things just last for
some time and according to my experience things like this last longer
than five days.  I would hold the bet that it takes at least seven days
and also this is dense ...
 
> > 4. I have copied my code of the Upload statistics into the repository.
> >    Please make yourself a bit comfortable with this - perhaps you see
> >    some chance for enhancements, but I think it is OK as it is.
> 
> Two reasons why I am not going to even _try_ improving this:
> 
>     i. You have far far more experience and knowledge with PostgreSQL than I do,
>     ii. goto i :D

Because I'm able to write SQL functions longer than one screen page? :-)
Well, you might just dive into this a bit anyway.  It might not harm to
understand it a bit - questions are welcome as always.
 
> > 5. I also would have loved to apply the measure Scott mentioned but
> >    I'm afraid we are a bit late.
> 
> Please refresh my memory. Which of Scott's ideas are we talking about?

H-index.

Just read
   http://lists.debian.org/debian-science/2011/01/msg00015.html
and following and perhaps something in our initial mail exchange for
GSoC project.
 
> > I tend to do the following summary of the work:  At the end of the GSoC
> > term we have not finished all the tasks we intended to do (finished in
> > the sense of thoroughly tested, stable and documented).  However, we had
> 
> I had consider the following complete:
> 
>     svnstat.py

It's probably complete but not tested with comparison to old values I
had (which is only possible in connection with gitstats because I mixed
them up via analysing the commit lists).

>     liststat.py

Yes, this gives reasonable results as we have seen in the teammetrics
list graph which I produced for our slides.  The comparison with the
other old graphs was not yet accomplished (my fault - I did not found
the time to finish this).  I assume it is OK.
 
> gitstat.py will of course be done before GSoC ends. And we can finish
> nntpstat.py also *just* in case.

Having working nntpstat.py as fallback is not bad, but as I said, proper
gitstat should have highest priority.  (In worst case I keep on using
liststat.py on the commit lists as this has proven to be working in the
past ... but we wanted to be better. :-))
 
> > You also had to spend some time in preparation of DebConf (uhmm, I would
> > hate begging for a Visa that much) and your time at DebConf has probably
> > also drained from pure coding time.  However, I really regard this as
> > much more important to become a member of the project rather than
> > writing some lines of code right in time for the and of GSoC term.
> 
> Total travel time + getting Visa + DebConf = 15 days. However, like
> you said, it was time well spent and I am really happy I made it to
> DebConf. Not only I got to meet you, but so many other wonderful
> people which was an experience in itself _plus_ which will ensure that
> I am going to be a part of Debian and contribute even after GSoC. So
> while 15 days could have been used for coding even more, but DebConf
> is DebConf and the 15 days will come again but not DebConf, so I am
> happy and not complaining!

IMHO 15 days well spend in your life ...
 
> > Moreover the BOF at DebConf has brought in some additional ideas which
> > are not coded as well - but it is easy to do it later.  (I just had some
> > quick look into UDD and I thinks we can do something about bugs closed
> > by a team member.)
> 
> This is perhaps the practical implementation of our project and once
> we have the metrics, doing this is going to be easy.

The problem here is:  We have the version that had closed a bug and we
know who uploaded this version.  Chances are good that the uploader is
the person who closed the bug - at least he has some share of the work
which was needed to close it.  However, it is not uncommon that an
upload has several changelog entries by different people and it might
happen not to seldom that a different person had submitted the relevant
change to close the bug.  I need to ask Enrico whether he has some
solution for this.

> > So in this sense I will fill in my mentor form in Melange that I will
> > not score full points for project finished as planed (which is in most
> > cases not realistic anyway) but that I regard the GSoC project as a
> > success regarding analysing team performance in Debian.
> 
> Heh, me too. Remember how I asked you and Scott that this project will
> be _under_ three months

Hehehe.  Poor young boy.  I'm coding since 1985 (starting with punching
cars) and I NEVER have seen a project finishing under the estimated
time.  :-)

> and now we were not able to actually complete
> everything? :) So many factors influenced that but overall I am happy
> (and you are!), so all is well.
> 
> I will finish Gitstat and we will see where to go from here.

Yep.

This project, the list and the repository will remain and evolve even
after GSoC 2011. :-)

Kind regards

      Andreas.

-- 
http://fam-tille.de