On Thu, Dec 2, 2010 at 12:29 PM, Gabor Grothendieck <ggrothendi...@gmail.com > wrote:
> On Thu, Dec 2, 2010 at 11:24 AM, Claudia Beleites <cbelei...@units.it> > wrote: > > On 12/02/2010 10:32 AM, Liviu Andronic wrote: > >> > >> Dear all > >> > >> On Wed, Dec 1, 2010 at 7:21 PM, Dominick Samperi<djsamp...@gmail.com> > >> wrote: > >>> > >>> The author line of the latest release of the R package > >>> Rcpp (0.8.9) was revised as follows: > >>> > >>> From: "based on code written during 2005 and 2006 by Dominick Samperi" > >>> > >>> To: "a small portion of the code is based on code written during 2005 > and > >>> 2006 by Dominick Samperi" > >>> > >>> From the info given in the thread, personally I'm sympathetic to > >> > >> Dominick's complaint: the latter message is no proper way to > >> acknowledge the original author of the package. As I see it, the > >> project either: > >> - explicitly mentions the original author and the active (current) > >> contributors (and perhaps previous ones), or > >> - lines up all previous contributors in a line and singles out the > >> active contributors > > > > - or in this case say that it was forked (when) from (Author)'s (package) > > (version) > > I think the danger in all this is that future developers might see > this discussion and then conclude that they would be better off > redeveloping existing packages encouraging a wasteful Not Invented > Here attitude rather than stand on the shoulders of others. That would > divert resources into nonproductive duplicative activities and slow > the growth of R. > > Perhaps the takeaway is (1) to be particularly careful about forking a > project and (2) also for package developers to try as hard as they can > to write their packages in a such a way that they can be added onto > externally rather than requiring modification of the package itself. > For example, DBI allows external database drivers and Rcmdr allows > external plugins. zoo can accommodate new classes of index without > modifying zoo itself. And of course R itself has specific facilities > for encouraging user contributed packages which do not require any > change at all to R itself. > > In fact, I wonder if its still not too late for the package in > question. Perhaps it would be possible to divide it into two packages > -- one would be the new code and the other would be the original base > package with just sufficient modifications to allow the new package to > consist of add-ons to it. (I haven't actually used the package in > question so I am not sure if this is realistic but thought I would > throw it out as a potential resolution.) > Actually, I attempted the reverse and created cxxPack (previosly known as RcppTemplate, and Rcpp before that), a package that depends on Rcpp (current version), and is designed to provide a kind of C++ application level interface on top of the lower-level interface provided by Rcpp, with particular focus on financial applications. For example, there is a C++ class ZooSeries that represents a zoo time series, and a C++ class DataFrame that represents an R data frame. These classes have C++ SEXP constructors, and the resulting object has no linkage to a corresponding object on the R side. In contrast, Rcpp tends to provide C++ classes that are wrappers for R objects. Objects of type ZooSeries and DataFrame can be mapped to the corresponding R objects and passed back to R, for example. Designing C++ classes that embed the logic of corresponding R objects is more difficult than just wrapping the corresponding R object, and I'm not sure this will pay dividends, but it is there nevertheless. There appears to be zero interest from the R community. It turns out that Rcpp (in its various incarnations) was always just a tool that I have used to build scientific applications, and it is unfortunate that issues related to this tool have overshadowed the real work. I surely do not want to waste my time merging cxxPack/Rcpp into yet another R/C++ solution, and I hope that the outcome of this discussion does not lead to this. Thanks, Dominick > > -- > Statistics & Software Consulting > GKX Group, GKX Associates Inc. > tel: 1-877-GKX-GROUP > email: ggrothendieck at gmail.com > > ______________________________________________ > R-devel@r-project.org mailing list > https://stat.ethz.ch/mailman/listinfo/r-devel > [[alternative HTML version deleted]] ______________________________________________ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
