[R] My remark on libraries

[Robert Gentleman]

Hi,
 This message is really informational -- some of us are trying to
 address these concerns -- the software is a few months away.

 First, I would like to second Ben's comments about the importance of
 both versioning and testing software, for this if for no other reason you
 should put your software into a package.

 Some of the distribution issues (and some of the dependency issues
 just mentioned on the MySQL thread) are being addressed. There is a
 (very alpha) version of reposTools at the Bioconductor site
 (www.bioconductor.org) that will make setting up your own website as an R
 repository (like CRAN) easier.

 In general, we will need to address some mechanism for attaching
 meta-data that is meaningful and searchable. This isn't easy nor is
 it a small project.

 Tools like install.packages and update.packages (-- our versions have
 a 2 as a suffix to avoid collisions), will try to resolve
 dependencies. There are a few other bells and whistles that need to
 be added -- but it seems to work most days.

 The Bioconductor mirror of CRAN has a suitable structure for trying
 this out (and there are some help files, that we try to keep in sync
 with the code base, but it is alpha)

 Robert


On Fri, Feb 07, 2003 at 09:01:04AM -0500, Ben Bolker wrote:
> 
>   This comes up from time to time; at the moment I can't find the thread
> in the mailing list archives.  The advantage, obviously, would be to make
> all this accumulated code available to the R community.  The disadvantage
> would be that such a heap of code would be highly heterogeneous -- some
> stuff would work, some wouldn't, some would work originally but obsolesce
> -- and the good stuff would get lost anyway.  On the other hand, you're
> right that it would be hard to find the functions one wanted in a set of
> "miscmisc" packages (but would it be any harder than finding it in a
> random heap of contributed code?)  
>   I guess I don't have a good answer; I would encourage you to bundle your
> stuff up into a little library, even if you don't think it's enough. (I am
> equally guilty of laziness: I've written a variety of packages for
> landscape generation, genetic stock analysis, 3D graphics, maximum
> likelihood ... but while I have put them up on my own web page at
> http://www.zoo.ufl.edu/bolker/R, I've never submitted them to CRAN for
> most of the reasons you cite.)
> 
> http://lark.cc.ukans.edu/~pauljohn/R/statsRus.html
> is a good collection of (tiny) bits and pieces, but it depends on a single 
> person for updating (and for keeping it coherent ...)
>   
>   Ben Bolker
> 
> 
> 
> On Fri, 7 Feb 2003, Eric Lecoutre wrote:
> 
> > 
> > Hi R community,
> > 
> > These days, I am writing some functions to work with 2-ways frequency 
> > tables ; you know all this tuff about measures of association: Chisq and 
> > derived (phi, cramer's v), tau b, tau c, somer's d and so on.
> > 
> > So I consider all those functions could be gathered in a single R file, as 
> > they are coherent and dealing with the same problem / objective (analysis 
> > of crosstables). Nervertheless, I wouldn't say there is here enough stuff 
> > to juystify the creation of a library. First, there is here few material, 
> > second it is not enough to cover the practical problem; which should be the 
> > goal of a library.
> > 
> > In fact, I have also lot of others splitted files with R code suitable for 
> > different purposes. And I could easily imagine I am not the only one among 
> > R programmers... And at the end, this is pity that I can't share this code 
> > (and BTW that I can't benefit of other's one!)
> > 
> > I though the solution was to deliver my own (eric)misc library, you know... 
> > But then we would rapidly be overcome by all miscmisc from world, and it 
> > would be difficult to find specific functions with all that libraries.
> > 
> > To me, a possible solution would be to have a page on CRAN listing such R 
> > files. Then, everyone would be able to upload R files with a short comment 
> > (what does the functions do). I am thinking about something less "strict" 
> > than libraries: for that, no reason to test the code (programmer's 
> > responsability). Then, maybe sometimes there would be enough pieces of code 
> > on a statistical subject to justify the creation of a library (still with 
> > the mind: a library is a collection of tools designed to help on a precise 
> > situation).
> > 
> > While I'm at it, it would also be nice to have a shared page on 
> > screenshots/graphics. For sure, I really enjoy to use R to produce 
> > wonderfull graphics. Not only would it be nice sometimes to show them, but 
> > it would also be the occasion to demonstrate R possibilities for newcomers 
> > (think on color graphics including mathematics). R would never have enough 
> > publicity!
> > 
> > Eric
> > 
> > 
> > __________________________________________________
> > 
> > Eric Lecoutre           Informaticien/Statisticien
> > Institut de Statistique                        UCL
> > 
> >                                (+32) (0)10 47 30 50
> >                             lecoutre at stat.ucl.ac.be
> >      http://www.stat.ucl.ac.be/ISpersonnel/lecoutre
> > __________________________________________________
> > Le vrai danger, ce n'est pas quand les ordinateurs
> > penseront comme des hommes, c'est quand les hommes
> > penseront comme des ordinateurs.     Sydney Harris
> > 
> > ______________________________________________
> > R-help at stat.math.ethz.ch mailing list
> > http://www.stat.math.ethz.ch/mailman/listinfo/r-help
> > 
> 
> -- 
> 318 Carr Hall                                bolker at zoo.ufl.edu
> Zoology Department, University of Florida    http://www.zoo.ufl.edu/bolker
> Box 118525                                   (ph)  352-392-5697
> Gainesville, FL 32611-8525                   (fax) 352-392-3704
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> http://www.stat.math.ethz.ch/mailman/listinfo/r-help

-- 
+---------------------------------------------------------------------------+
| Robert Gentleman                 phone : (617) 632-5250                   |
| Associate Professor              fax:   (617)  632-2444                   |
| Department of Biostatistics      office: M1B20
| Harvard School of Public Health  email: rgentlem at jimmy.dfci.harvard.edu   |
+---------------------------------------------------------------------------+