[pkg-bioc] Re: Google SoC R project

[Dirk Eddelbuettel]

Chris,

On 21 April 2006 at 23:54, Christian Bird wrote:
| Thanks a bunch for the info.  I actually know Duncan and have met with
| him a number of times discussing my research (we're datamining open
| source projects (mailinglists, cvs, etc.) and analyzing the data in
| R).  I'll take a look at the sites you mentioned for more info.  I

Try checking out the script from the alioth project, eg via

http://cvs.alioth.debian.org/cgi-bin/cvsweb.cgi/tools/?cvsroot=pkg-bioc

The more current one is the 

just do see what it does:
-- use R to get installed and available packages
-- use stupid static data to map R package names to Debian package names
-- in several loop iterations, for each package
	* see if it has not been built yet, but needs building
	* parse R's DESCRIPTION, write debian/control, debian/copyright,
	  the standard debian/rules etc
	* build the package
	* continue the loop

So it could do with better dependency checking, better copyright status
analysis, ... and generally a re-thought.  And generalisation: from CRAN to
BioC to maybe other repos (Omegahat, RoSuDA [ Simon Urbanek et al at Anthony
Unwin's research group ], ...)

| figure that if I have a decent game plan and mention concrete plans in
| the proposal, I have a better chance of it getting accepted by google.
|  If not, I still think it's worthwhile.  I don't think that getting
| this done would necessarily directly further my research, but
| indirectly it will help me be more productive later on.  Have a good
| weekend.

Absolutely. Feel free to run questions and proposal by me / the list (which I
added back to the CC).

If you want to submit this via Debian as a sponsoring institution, we need to
work on it a little anyway. 

Best regards, Dirk

| 
| -- Chris
| 
| On 4/21/06, Dirk Eddelbuettel <edd at debian.org> wrote:
| >
| > Hi Chris,
| >
| > You just caught me before going to bed :)
| >
| > On 21 April 2006 at 21:10, Christian Bird wrote:
| > | Dirk,
| > |
| > | I'm interested in applying for Google's summer of code and
| > | implementing a debian package generator for CRAN packages.  I'm a C.S.
| > | Ph.D. student at UC Davis and I use both R and python quite a bit for
| >
| > [ Off-topic: Do you know that R Core member Duncan Temple-Lang is now at UC
| > Davis, in the stats dept?  I guess you probably do ... ]
| >
| > | my research (and of course, use debian).  I saw automatic package
| > | generation of CRAN packages as a suggestion and it's something that
| > | would directly help me in my work so I'm interested in it.  I have two
| > | questions for you.  First, have many other people contacted you about
| >
| > Nope, you're number one. Only posted two days ago though...
| >
| > | working on this?  Second, is it possible to get a bit more detail
| > | regarding what you're wanting exactly?  Is it just automatic packaging
| > | of the packages or is there more to the equation (regression testing,
| > | etc.).  Thanks for any info you can provide.
| >
| > The page(s) on alioth and the corresponding mailing list archives may help
| > you.
| >
| > In essence, several years Albrecht Gebhard ago wrote a perl hack to build
| > SuSE package off CRAN. That later landed in Detlef Steuer's lap as Albrecht
| > moved on to Debian. Maybe two years ago I grabbed that script and turned it
| > into a Debian hack -- and generated packages for my Quantian distro. I then
| > joined forces with Matt who had started an alioth project for BioC.  The BioC
| > / CRAN project is still on alioth (and you could join that _now_ irrespective
| > of what happens with SoC, get an alioth id and I will add you), and a few
| > good folks have sinced joined and pushed the script further. It is in version
| > control and the wiki pages have some notes on current statis.
| >
| > It is still a pretty simple Perl hack. Given how easily CRAN packages build
| > (given how regularised they are with their own QA and unit-like testing...)
| > it is mostly a matter of mapping Depends etc. And polishing so that it runs
| > daily in an automated manner.
| >
| > I also do not know how likely we are to get this approved a) within Debian
| > and b) with Google. I *did* learn that the R Core group had been approached
| > (presumably directly by Google) so we may have a chance with a proposal.
| >
| > Not sure if we can flesh this out so that it maps to your PhD research.  But
| > we can definitely do _something_.  I have meant for so long to create a
| > process that follows daily CRAN updates, builds new debs, summarizes changes
| > to a mailing list (or rss feed) etc.  Not hard -- but I just haven't found
| > the time with work, 70+ Debian packages, Quantian ...
| >
| > I'll see this to the pkg-bioc list where I had dropped a note about the SoC
| > idea.
| >
| > Cheers, Dirk
| >
| > |
| > | -- Chris
| > |
| > | --
| > | Christian Bird
| > | cabird at gmail.com
| >
| > --
| > Hell, there are no rules here - we're trying to accomplish something.
| >                                                   -- Thomas A. Edison
| >
| 
| 
| --
| Christian Bird
| cabird at gmail.com

-- 
Hell, there are no rules here - we're trying to accomplish something. 
                                                  -- Thomas A. Edison