Re: Fun with the BTS and an SQL DB
On 05/05/07 at 23:34 +0200, Stefano Zacchiroli wrote:
> On Sat, May 05, 2007 at 12:39:56PM +0200, Holger Levsen wrote:
> > Wow. Only 1000 packages have more than 10 bugs and more than 3000 packages are
> > bug-free (if you counted source packages). Of course, we still can & should
> > do better :-)
>
> On Sat, May 05, 2007 at 08:27:16PM +0200, Lucas Nussbaum wrote:
> > Yes, I was also surprised by the results. I'll try to generate some nice
> > graphs with all those figures.
>
> Actually I'm not that surprised. I'm thinking under the assumption that
> *a lot* of debian packages are de facto orphaned, meaning that they were
> packaged by people interested in having their toy packages in debian,
> but which are actually not used by anyone else (and hence no bug has
> been reported against them).
Well, I'm not sure that we have so many "pet packages". See
http://blop.info/bazaar/ecdf-popcon.png (generated before FOSDEM, but
unlikely to have changed a lot since) : a lot of packages don't have a
lot of installations (~85% with <1000 installations, 50% with <100, ~10%
with <10).
> A way to check if I'm right of (hopefully) wrong would be to "weight"
> the result using popcon results or even simply removing from the count
> packages whose poocon installation count is under a given threshold
> (to be determined ... do not ask me how :)).
A better solution could be to compute the "number of bugs per popcon
install" ratio for each package, and see if some packages have totally
abnormal results, or better, to graph the couples (number of bugs,
popcon installs) and then see if the graph looks like something one can
interpret. But my statistics skills are a bit rusty, so help is
welcomed!
> I haven't followed the entire thread, are you able to join the BTS
> data with popcon results using the current database?
not yet. But importing the popcon results in a table is easy, and then,
it's also easy to compute one of the above data. However, I'd like to
concentrate on my initial goal first, that is, be able to store
historical data about number of bugs.
> BTW, thanks for your work! Cheers.
You're welcomed :)
--
| Lucas Nussbaum
| lucas@lucas-nussbaum.net http://www.lucas-nussbaum.net/ |
| jabber: lucas@nussbaum.fr GPG: 1024D/023B3F4F |
Reply to: