Recently, I have started to rewrite some of the Perl scripts I used for stemmatological experiments (a few of them briefly commented on this site, the others in a to be published article) in R. Even if R is a lot slower than Perl (at least in that very particular case), I find it has some crucial advantages, and makes the scripts and procedures :
- easier to exchange with other researchers (that is, easier for them to install, use, benchmark, and the like) and as such good for results reproducibility ;
- part of a larger framework of statistical methods, and so facilitates the implementation of similar methods, the testing of various approaches over the same data, serendipity, etc.
These seem so convincing that I started wondering a few days back if it wouldn’t be of interest to integrate those functions into a package (in that regard, the ability to integrate sample datasets, such as the ones I used in my experiments, in the package, is very decisive for point 1), and, why not, make this package a broader one, able to integrate other stemmatological methods. To my knowledge, there is no stemmatology package as such in development, though several methods used in stemmatology are probably already implemented in R, and, also, the Stemweb project uses R, so they probably have already done so as well. I’ll continue converting the scripts and see how that goes, but I would very much welcome at any point any remark, commentary, suggestion or proposition regarding this potential package.
Here is an interesting slideshow by Rory Winston (2011), Creating R packages : <https://fr.slideshare.net/rorywinston/creating-r-packages> (and the filmed version here).
Another informative reading is Spencer Graves & Sundar Dorai-Raj (2008), Creating R Packages, Using CRAN, R-Forge, And Local R Archive Networks And Subversion (SVN) Repositories, <http://cran.r-project.org/doc/contrib/Graves+DoraiRaj-RPackageDevelopment.pdf>.
On a final, undirectly related, note, I would advise anybody interested to have a look at the exciting developments in the field of stemmatological software and particularly at what is done concerning the Stemweb and Stemmaweb apps :
Tara Andrews, Simo Linkola, Teemu Roos, Joris van Zundert, « White Paper: Interoperability between Stemmatological Microservices », The Tree of Texts, <http://treeoftexts.arts.kuleuven.be/?p=58>, 7 mai 2013.
As an addendum to this post, I have started a question about the reasons to create (or not) an R package, on the CrossValidated StackExchange site :
Why and when create a R package?
Ping : Roadmap for an R Stemmatology Package | Sacré Gr@@l