[Popcon-developers] Bug#291747: popularity-contest: Please give stats on source packages as well

[Bill Allombert]

On Tue, Jan 25, 2005 at 08:52:21PM +0100, Martin Quinson wrote:
> Yes, what is important to me is that we rank all the source packages. I
> would remove from the ranking the "not in sid" package (or move them to the
> end of the ranking). It's at least the easiest to do.
> 
> > 2) How do you define the popularity of source packages ? 
> > As the sum of the popularity of each binaries packages ?
> > This is what is used for Maintainers ranking currently.
> 
> It makes sense for maintainers, but I would say that for source package, we
> should use the max of all achieved ranking. I'd say that it's the metric
> which would help the translators looking for which package to translate. I'm
> not completely sure, in fact. Doing 2 rankings is also easy, but I'm not
> sure it's worthing our time.

Hello, I have made a preview here:
<http://people.debian.org/~ballombe/popcon-exp>
Please tell me if anything is broken.

It use the sum. I could easily change it to use the max, however the
correct way to compute the max would be to read popcon reports directly
and to count the number of thoses that report at least one package
provided by the source package (iow, to merge the reports).

suppose we have a source package foo generating foo-bar and foo-dev,
an 2 users.

user1 report: foo-bar
user2 report: foo-dev

In the all-popcon-result file (anonymized result) we have:
foo-bar: 1
foo-dev: 1

So if we use the sum we get foo:2, and if we use the max we get foo: 1.
However, looking directly at popcon reports report we get:
user1 report: foo
user2 report: foo

so we get 2 reports for foo, which seems more correct since there are
effectively 2 users of the foo source package.

OTOH, if we had
user1 report: foo-bar foo-dev
user2 report: 

We would get 2 for sum and max, but in fact only one user use foo.

The problem is that the website is build from the anonymized results
for confidentiality. To get correct result here, I would need to add
source packages info to the anonymized results files, and I am not
to keen changing the format of that file.

Cheers,
-- 
Bill. <ballombe@debian.org>

Imagine a large red swirl here.