Re: Packaging r-bioc-simpleaffy
On 02/26/2011 12:09 AM, Tony Travis wrote:
> On 25/02/11 19:55, Manuel Prinz wrote:
>> On Tue, Feb 22, 2011 at 10:34:41PM +0000, Tony Travis wrote:
>>> If you want me to name some important Bioconductor packages I use
>>> then I suggest "simpleaffy" and its dependencies:
>> Out of curiosity, I hacked a script (quick and dirty) that gathers
>> dependancy information from the Packages file of Steffen's cran2deb
>> repository and draws a dependency graph that can be processed
>> with dot. An example output is attached. It also gives some stats
>> on stderr about how many packages depend on a package. Dunno whether
>> or not this is useful, though.
> Hi, Manuel.
> Well, I think that it's very useful - Thanks!
>> Dashed entries are BioConductor packages, solid ones are from CRAN.
>>> From the later, only r-cran-rsqlite is not yet in Debian. In order
>> to package r-bioc-simpleaffy, one CRAN package and 11 BioConductor
>> packages need to be packaged. One could start from the source
>> packages created by cran2deb. I'm not a user of BioConductor myself
>> but if there are a few people willing to work on that, I'd volunteer
>> to help with reviews and uploads. The packages should be beginner
>> friendly, so maybe someone would like to pick those up?!
> My original idea was to re-package the Bioconductor packages that are
> required by GenePattern for Debian/Ubuntu. This would require quite a
> lot of packages, and their dependencies, to be re-packaged. However, now
> that we have a way of automatically creating source packages with
> cran2deb it might be a tractable proposition to prepare the sub-set of
> Bioconductor packages needed for GenePattern for submission to Debian.
> My objective is to package GenePattern for Debian/Ubuntu, using the
> existing CRAN packages in Debian/Ubuntu to satisfy dependencies. I'm
> still waiting for the Broad Institute to tell us under what licence we
> can redistribute GenePattern, but they are interested in our efforts.
Some of those packages are updated rather frequently. This is where
the main work goes into - not into the initial packaging, and this is
why this cran2deb bits came into play.
The philosophy of Dirk and Charles is that the automated builds shall
substitute the manual builds. If there is something to be changed to
the build process, then this patch should become part of cran2deb
(which can be done) and/or be communicated back upstream. This is why
the Debian version information does not offer any "~" magic and does
not do the "0cranX" for new packages like the friendly Ubuntu does.
We could overrule this, certainly, as Dirk keeps reminding us, this
is our repository. But we certainly need to think twice about if those
packages are not possibly too much evolving to be of any use in the
main distribution in the first place. For BioConductor's Debian packages
I would tend to think that a separate repository may be preferable.
Feedback on the usability of those packages so far was rather limited.
We shall now announce it to Debian's R SIG mailing list, give it some
time there and then go to the BioConductor mailing list.