Show most active projects on cloud, rather than most alphabetical

[Jonathan Lange]

Currently the cloud on code.launchpad.net is limited to a few projects. The limit is calculated by taking the first N projects alphabetically, rather than the first N projects by activity. The cloud would be more useful if it were the most active projects.

There are two possible negatives:
    * this would be event slower than the current cloud calculation
    * the new cloud would lack the visual variety of the current cloud

[Jonathan Lange (jml)]

An alternative would be to randomly sample the projects that make up the front page cloud.

[Robert Collins (lifeless)]

I don't see a good reason for it to be slower; if anything, doing one query would be faster :P. As for the variety, having decay on activity, and new projects popping up, would be a massive improvement over the current completely-static-except-for-registered-low-alpha-order projects.... which is why I asked about this in IRC :)

[Martin Albisetti (beuno)]

On Thu, Feb 5, 2009 at 10:25 PM, Jonathan Lange <email address hidden> wrote:
> Currently the cloud on code.launchpad.net is limited to a few projects.
> The limit is calculated by taking the first N projects alphabetically,
> rather than the first N projects by activity. The cloud would be more
> useful if it were the most active projects.
>
> There are two possible negatives:
> * this would be event slower than the current cloud calculation

Even if you sort by the top 30 and limit them?

> * the new cloud would lack the visual variety of the current cloud

Not if they where relative and not absolute. If the project with less
commits is the base for 0, and the projects with most commit is 100,
than the visual result would be the same.

--
Martin

[Robert Collins (lifeless)]

Martin, I don't understand your comment. Visual variety doesn't depend on sampling from least inactive->most active, it depends on the set shown changing over time, Currently the change is 'new project starting with 'A''. Nearly anything will offer more variety :P.

More usefully though, grabbing the top N (say 15) active projects should be fine, unless those 15 stay permanently most active.

Perhaps you mean though, that activity should be relative to code size? (With commits a placeholder for code size?) So 1 commit in a project with 10 commits total is 10% activity, and 1 commit in a project with 50000 is 0.002% ? Or something like that?

[Jonathan Lange (jml)]

On Tue, Feb 10, 2009 at 11:24 AM, Robert Collins
<email address hidden> wrote:
> Martin, I don't understand your comment. Visual variety doesn't depend
> on sampling from least inactive->most active, it depends on the set
> shown changing over time, Currently the change is 'new project starting
> with 'A''. Nearly anything will offer more variety :P.
>

Martin is answering the question that I intended to ask. You answered
an interesting question that I didn't mean to ask. :)

[Martin Albisetti (beuno)]

On Mon, Feb 9, 2009 at 10:24 PM, Robert Collins
<email address hidden> wrote:
> Martin, I don't understand your comment. Visual variety doesn't depend
> on sampling from least inactive->most active, it depends on the set
> shown changing over time, Currently the change is 'new project starting
> with 'A''. Nearly anything will offer more variety :P.
>
> More usefully though, grabbing the top N (say 15) active projects should
> be fine, unless those 15 stay permanently most active.
>
> Perhaps you mean though, that activity should be relative to code size?
> (With commits a placeholder for code size?) So 1 commit in a project
> with 10 commits total is 10% activity, and 1 commit in a project with
> 50000 is 0.002% ? Or something like that?

I think I meant that you would sample the top X, and that the lowest
activity would be 0, and the the max would be 100, and you adjust
accordingly, no matter how much activity it is, just relative to the
lowest one.
I'm not 100% sure that it's the best approach, as the least active
could be quite active, and give out the impression it's not, but it
sure is a way to avoid having everything be very big on the page :)

--
Martin