Forums : Feedback Forum

First off, thanks to all of the ohloh team -- this is an excellent resource. We're having fun it. There have been several issues noticed, however, that I hope will be looked into at some point.

The biggest issue I noticed is that the processing doesn't seem to take Attic/deleted/moved files into account at least when computing the years contributions of developers. Our project, BRL-CAD -- http://www.ohloh.net/projects/3996 -- has almost 25 years of CVS history (potentially one of the longest retained project histories in revision control system) but the counts on our project page don't reflect this. For example, it list the project's original author (mike) at only 7.0 years contribution when that should be 17 years. There are commits from him all the way back to 1983 through 2000.

One possible cause for the 7.0 year contribution is that in 2004, our repository was vastly reorganized changing from a top-level heavy repository to a more hierarchical organization. This resulted in a ton of files and directories moving around in CVS (via delete/readd). If our repository is checked out unpruned or even if Attic files for existing directories were considered, the logs are massively more extensive. Since our reoganization was so extensive, the years contribution is pretty much wrong for all but the newest contributors (since we went open source in 2004) which is no noticeable given the major discrepancy for all our devs. If it is taking the Attic into account .. then I can only imagine there is some other bug/assumption in the years contribution calculation.

Example statcvs report that shows our reorganization and extent of history: http://ftp.brlcad.org/statcvs/

Speaking of statcvs.. one of the coolest aspects of the statcvs report is their authorship speckle graph.. that would make for a great ohloh feature, especially if it took the magnitude of the commit into account (which statcvs doesn't). Example, red graph near bottom: http://ftp.brlcad.org/statcvs/authors.html

Other feature requests (not that you probably don't have more than enough to do already):

• Include BSD-style licenses in Licenses section file count
• Allow specification of exclusion paths to not process (e.g. our src/other is entirely 3rd party dependency codes, shouldn't be included in our statistics)
• Account for Attic changes in project costs
• Update enlistment reprocessing more frequently (perhaps once a week or at least on request)
• Add an age of project revision history metric (even our non-attic files go back almost 20 years)

Thanks again for the great site. Look forward to seeing more improvements!

Cheers!
Sean

sean over 17 years ago

Hi sean,

Thanks for your very interesting feedback. It's nice to get very detailed bug reports.

You have a ton of cool ideas and links in this post, so I want to spend some time thinking about them before I comment, but in the short term I want to address the years of contribution issue.

I think the key here may simply be that it's not obvious what we mean by 7 years contribution, and our terminology is inconsitent on different pages.

Our source control analysis does indeed stretch back the full 23 years to December of 1983, which you can see on the commits listing. Incidentally, I think you may hold the record for longest continuous history on our system -- I'll have to check.

The reason we say that mike has only 7.0 years of contribution is because we count only months in which mike actually checked in code. If you look at mike's timeline, you can see that before 1995 he was only checking in code during one or two months out of the year. In total, he checked in code during 84 different months, for 7.0 total years.

Very soon, we plan to expand our developer analysis to include factoids such as Joined the team in 1983 and Retired from the team in 2000 which might clear up some confusion, and I concede that we should be much more clear about what we mean when we say 7.0 years contribution.

I'll study the rest of your post in detail and get back to you again. Thanks again for the great feedback,

Robin

Robin Luckey over 17 years ago

Robin,

Thanks for the quick reply and reference to the commits listing. I do see now that there are a speckling of Attic files being counted from that far back, but it is missing a plethora of commits. For example, the commits listing shows exactly one commit for 1984 (second to last page). However, if I ask CVS for history on just one file:

.. [ from an existing unpruned checkout ] ..
~/brlcad morrison$ cvs log rt/main.c | grep 1984
date: 1984/11/30 04:04:13; author: mike; state: Exp; lines: +12 -12
date: 1984/11/29 07:04:33; author: mike; state: Exp; lines: +6 -6
date: 1984/11/27 06:59:47; author: mike; state: Exp; lines: +58 -32
... [ trimmed ] ...
~/brlcad morrison$ cvs log rt/main.c | grep 1984 | wc
27 267 1881

There's 27 commits in 1984 on just that one file from mike. There are veritably hundreds of files like that with commit traffic missing for most of the 80's and some of the 90's, and that span several authors' activity. It's not clear to me what the correlation is where brlcad/jove/ would be read but not (any of) the files in brlcad/rt/ nor those in a couple dozen other directories that are similarly in the Attic. Hopefully it might give you a lead on where to look though.

Thanks again for the reply and great work. The developer analysis idea you mentioned sounds like a great feature to say the least. :-)

Cheers!
Sean

[edited due to formatting issues, apologies on all the mods.. a preview option would be nice.. ;-)]

sean over 17 years ago