On 08/06/2009 05:53 PM, Michael DeHaan wrote:
What I'm thinking so far:
- Parallelize computations on individual people
- Generate reports based on files (new, in addition to people reports)
- save CSV to actual CSV file named after the repo
- consider aggregrate human stats across projects
- sort the CSV just for nicer output
- allow projects to scan to be listed in a config file (JSON)
- output graphs using OSS graph tools
- convert Date Strings into Date objects for better comparisons
- generate more stats on a per person basis
- list of times between commits
- average time and stdev
- standard deviation of lines added/removed (impact)
- roll up reports on a per project basis
- report on how powerful is the long tail
- avg commit size / stdev
- avg commits month / stdev
- burnout/superstar indicator (rate of change)
- what is the acceleration of the rate of time between commits
- what is the acceleration of the size of commits
- use aggregrate project data to produce comparisons
- X/Y graph of various projects -- commit volume vs count (OR:
things like #avg commit volume vs # contributors)
This all seems reasonably easy to achieve and we have already mined
most of the data, it's mostly about massaging it now.
Ah, wrong list. Well, might as well bring this up here.
I'm working on stage 2 of the ideas behind EKG -- source code data.
It will be unaffiliated with EKG, but I'm trying to mine some data that
will tell us meaningful things -- and things that ohloh and github don't
currently report.
Ideas welcome.
More info:
http://michaeldehaan.net/2009/07/24/git-python-and-perhaps-the-start-of-s...
(note: I'm /not/ using git-python)
http://michaeldehaan.net/2009/08/04/git-foss-stats/
http://michaeldehaan.net/2009/08/06/dear-lazyweb-favorite-open-source-cha...
--Michael