Re: PopCon (was Re: Drop 'epic' package? (was incorrectly Re: ITA: epic4))

[Richard Braakman <dark@debian.org>]

On Tue, Feb 20, 2001 at 09:35:08AM -0800, Rick Younie wrote:
> This is the chunk of python that parses the results.
> 
>     def vote_for(vote, package, entry):
>         now = time.time()
>         if entry.atime == 0:  # no atime: empty package
>             vote.empty_package = vote.empty_package + 1
>         elif now - entry.atime > 30 * 24*3600:  # 30 days since last use: old
>             vote.old_unused = vote.old_unused + 1
>         elif now - entry.ctime < 30 * 24* 3600 \
>           and entry.atime - entry.ctime < 24*3600:  # upgraded too recently
>             vote.too_recent = vote.too_recent + 1
>         else:                   # otherwise, vote for this package
>             vote.yes = vote.yes + 1
> 
> I make this,
> 
>  Old = hasn't been used in 30 days
>  Recent = been used in the last 24 hours but was upgraded/installed
>           less than 30 days ago

I think you misread this.  The 24 hour condition compares atime to ctime.

   Recent = was upgraded/installed less than 30 days ago, and has
            not been used since the day of installation

>  Vote = was used in the last 24 hours and installed/upgraded > 30 days ago
>         or used > 24 hrs ago and installed < 30 days ago

   Vote = was used less than 30 days ago, and more than one day after
   	  installation

> 
> So (Vote + Recent) compared to Old gives a good idea of 'popularity' IMO.

I don't think so.  That doesn't count people who remove the package
after they stop using it.  And I think Recent should be taken as a
measure of inaccuracy.

Richard Braakman