Here is the exchange I had with upstream. Apparently one should consider the C interface and the R interface as two /completely/ different product. For instance the R interface replace some deep down call (rand()) with something else.
---------- Forwarded message ---------- From: Tamas Nepusz <ta...@cs.rhul.ac.uk> Date: Thu, Sep 17, 2009 at 2:47 PM Subject: Fwd: Re: igraph / R-interface To: mathieu.malate...@gmail.com Cc: csardi.ga...@gmail.com Hi Gabor, I think you meant to cc the forwarded message below to Mathieu as well. Yes, you're right -- if igraph in R uses the random number generator of R, then it's not feasible to make it depend on libigraph, it's best to keep them separate. I wasn't aware of that. -- Tamas ---------- Forwarded message ---------- From: "Gábor Csárdi" <csardi.ga...@gmail.com> To: Tamas Nepusz <ta...@cs.rhul.ac.uk> Date: Thu, 17 Sep 2009 14:31:45 +0200 Subject: Re: igraph / R-interface Tamas, Mathieu, On Wed, Sep 16, 2009 at 6:29 PM, Tamas Nepusz <ta...@cs.rhul.ac.uk> wrote: > Hi Mathieu, > >> What is the interfaces/R in igraph from sf.net then ? > It depends on which file did you download from sf.net. > There are two files you could have downloaded: > > 1. The source code of the C library: > http://switch.dl.sourceforge.net/sourceforge/igraph/igraph-0.5.2.tar.gz > This might contain some files in interfaces/R, but I guess the > majority of them is not included. In particular, if your > interfaces/R/src subdir does not contain a file called > rinterface.c.in or rinterface.c, then you don't have the source code > for the R interface. > > 2. There is also the source code of the R interface: > http://master.dl.sourceforge.net/sourceforge/igraph/igraph_0.5.2-2.tar.gz > For reasons unknown to me, the source code of the C core is included > in the downloadable tarball for the R interface -- I guess the reason > is that this is the only way to enable R automatically download and > install igraph when you type "R CMD INSTALL igraph" from the command > line. The goal is to make it easy to install the igraph R package. That's why it is self-contained. This way you can install it in ten seconds on any supported system. > So if you are working with the tarball I mentioned in point 1), that's > no good for compiling the R interface. For the R interface, you need the > tarball mentioned in 2) and you have to tweak it somehow to link to an > already compiled igraph dynamic library instead of compiling igraph > again and linking to it statically. I don't know how other R packages > requiring external libraries get around this problem. Why tweak it? Why isn't the included library good? Anyway, tweaking it is not a good solution for several reasons. One is, that igraph uses different code when compiled as an R package, e.g. it uses R's random number generator. So the dynamic library is the R package and the "general" igraph dynamic library are not the same. Think of these (I mean libigraph and R-igraph) as two products that provide related functionality, and probably the best is to package them (if this is the intention) independently, i.e. without R-cran-igraph (or whatever its name would be) depending on libigraph. Best, Gabor > -- > Tamas > > -- Gabor Csardi <gabor.csa...@unil.ch> UNIL DGM << Email has been scanned for viruses by Altman Technologies' email management service - www.altman.co.uk/emailsystems >> -- Mathieu -- To UNSUBSCRIBE, email to debian-med-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
