Gentoo User Statistics -- what packages have you installed?

Message

kineticbeing · Post by **kineticbeing** » Fri Feb 13, 2009 8:06 pm

Is anything happening with this?

g-user · Post by **g-user** » Wed Apr 08, 2009 3:29 pm

You don't think make special ebuild for obtaining stats like that?
For example Debian project use popularity-contest package for obtain information like that and ask about this at install steps. This may give more information.

oldefortran · Post by **oldefortran** » Fri May 22, 2009 6:47 pm

Would it not be more interesting if what programs were run were collected?
It could be quite intrusive, but, I think, interesting.

How to get accurate stats, I have no idea...

vincent- · Post by **vincent-** » Wed Jun 24, 2009 9:21 am

I have seen in the web

Code: Select all

We now have 1834 entries.

Then, I have made this:

Code: Select all

curl -d "arch=$(uname -m)&packages=$(qlist -CI)" http://stats.larrythecow.org/
Thanks :)  We now have 1834 entries.

Must be 1835.

Then I have reload the web page and says:

Code: Select all

We now have 1835 entries.

FastTurtle · Post by **FastTurtle** » Mon Jul 13, 2009 3:33 pm

Couldn't get hold of the stats page to upload my packages as the website is not responding. Is it down or gone?

Mike Hunt · Post by **Mike Hunt** » Mon Jul 27, 2009 12:00 am

Thanks

We now have 1858 entries.

pa4wdh · Post by **pa4wdh** » Mon Jul 27, 2009 10:45 am

Hi,

Do you have some plans of actually using the information to generate some stats or whatever ?

The counter is at 1860 now, so i guess that's a good starting point.

Best regards,
pa4wdh

micr0c0sm · Post by **micr0c0sm** » Fri Aug 28, 2009 7:19 am

the world file would be much more useful imo.

V-Li · Post by **V-Li** » Fri Aug 28, 2009 9:16 am

There has been a Summer of Code project to use Smolt for that...it got quite far and sping is still working on it.

ToeiRei · Post by **ToeiRei** » Fri Aug 28, 2009 9:51 am

Thanks We now have 1870 entries.

The result would be quite interesting...

Rei

Roman_Gruber · Post by **Roman_Gruber** » Fri Aug 28, 2009 10:15 am

Code: Select all

r2d2 / # curl -d "arch=$(uname -m)&packages=$(eix -I --only-names)" http://stats.larrythecow.org/
Thanks :)  We now have 1872 entries.

lol libaries rules!!! But they are important. I think also gcc, kernels ...

Well to filter the results is another topic, right? He wantet, what users have installed !

V-Li · Post by **V-Li** » Thu Sep 03, 2009 8:18 am

As an instruction:

Use layman to get the sping overlay.
Unmask app-admin/gentoo-smolt
Emerge gentoo-smolt
Generate a UUID with cat /proc/sys/kernel/random/uuid > /etc/smolt/hw-uuid
Run smoltSendProfile --server=http://smolt.hartwork.org:45678/

You will see the statistics on http://smolt.hartwork.org:45678/static/ ... entoo.html. This is still in testing phase, but it works quite well.

EDIT: Added generation of UUID to list.

ArneBab · Post by **ArneBab** » Mon Sep 07, 2009 7:56 pm

As a short explanation to the above: That's currently the definite way to collect stats on Gentoo, so if you submitted to the original stats server, please do so here, too!

I already did, and this is my profile:
- http://smolt.hartwork.org:45678/client/ ... 85a610b3d1

Kollin · Post by **Kollin** » Mon Sep 07, 2009 8:58 pm

Mine:

http://smolt.hartwork.org:45678/client/ ... 8411601d42

Post by **sping** » Tue Sep 08, 2009 2:30 am

AllenJB wrote:
Top 10 packages in each category.

Specific package versions that are not used by anybody (or by only a handful - less than 20 for example)

Popularity over time graphs for each package

Perhaps for each category too - identify which appears to be the most popular desktop environment for example

Created "bugs" for these four:

nightmorph wrote:Also, I'd like to see the number of users who have packages installed that are not from the Portage tree. Furthermore, of that group, what percentage of those packages are not available in the tree? It'd be interesting to see how many 3rd party resources, such as overlays, upstream ebuilds, or homebrewed standalone ebuilds, are in use.

Does the report table on repositories answer that to your contentment?

ArneBab wrote:Relations between packages: Which packages are installed together with which other packages most of the time?

With about 13,000 packages and a growing number of submissions that would strongly increase report generation time. I wonder if the resulting data would be any helpful as there will be high correlations to all sorts of unrelated packages, e.g. Amarok will have high correlations with Vim.

alistair wrote:2) If you have package version, which ones have outstanding GLSA's ( Tho it would be very inaccurate so maybe we should leave this one alone).

What dio you mean by "outstanding" GLSA? If there was something like repeatedly-updated XML file with all package names and versions that are affected by some sort of vulnerability integration into the report (e.g. marking packages red or linking to GLSAs) would be possible. Is there such thing?

alistair wrote:3) Maybe find all text editors and then have stats based on them, ie count, whether there are relationship between text editors. eg. if there a relatonship betwwen system with emacs also having kwrite ( or kate and vim ), etc ,etc

A table dedicated to text editors would be no problem. I wonder if the correlation would be worth the effort. Convince me

node_one wrote:Bug statistics for the popular packages, maybe?

Please elaborate on that.

Sebastian

ArneBab · Post by **ArneBab** » Tue Sep 08, 2009 5:37 am

sping wrote:
ArneBab wrote:Relations between packages: Which packages are installed together with which other packages most of the time?
With about 13,000 packages and a growing number of submissions that would strongly increase report generation time. I wonder if the resulting data would be any helpful as there will be high correlations to all sorts of unrelated packages, e.g. Amarok will have high correlations with Vim.

I wouldn't make that the default part, but rather something people can generate _by hand_, if they want to create a news article or similar.

I assume that doing correllation calculations would take quite a lot of processing power, so maybe people could just say "smolt, I need a 'table' of all installed programs per user".

Then they can do their own calculations (with their own processing time).

If you want to make it easy for people that would be something like

Code: Select all

smolt processData --server=... --correllation="items:installed_packages,subjects=users"

(with improved option names)

Gentoo User Statistics -- what packages have you installed?

Larry the Cow Dead/MIA