Re: Re: igraph / R-interface
On Thu, Sep 17, 2009 at 4:32 PM, Mathieu Malaterre
<mathieu.malaterre@gmail.com> wrote:
> On Thu, Sep 17, 2009 at 4:01 PM, Gábor Csárdi <csardi.gabor@gmail.com> wrote:
>>>> Why tweak it? Why isn't the included library good?
>>>
>>> Because I use libigraph with external blas / lapack and lots of
>>> options which are not the one used in the r interface.
>>
>> You can build the shared library in the R package with external
>> blas/lapack, and your lots of options, as well.
>
> I could not figure that out, there were no instructions on how to do that...
Just replace any option you want in the 'Makevars' file.
See the 'Writing R extensions' document.
[...]
> Now as a debian packager, as I have to explain why I am polluting
> debian disk space with two packages which differ with a single ifdef
> on the c-rand() versus the r-rand() function.
>
> This is not clear to me why the rand is so burried down in the igraph
> library. The rand function should be a parameter, so that the
> R-interface would use the r-rand() function, the python-interface
> would use whatever python-rand function is, java would use
> whatever-rand function it needs. But I am not going to triplicate the
> source code of igraph for every single target language:
You can use the same source code, just like we do. There is just one
bzr tree for all the stuff, not a single file is duplicated in it.
Some of the binaries will be duplicated, though, at least until we
implement the hooks I mentioned in the previous email.
But actually, I am not convinced that having R-igraph in Debian would
be a big benefit. I've used Debian and Debian-based distributions for
more then ten years now, but always installed R packages from CRAN and
not from Debian. They provide the same flexibility, CRAN has more
packages, it is up-to-date and can be used without admin rights, from
R itself.
So I would propose to package the C library only, and wait until we
"fix" the R package.
[...]
G.
--
Gabor Csardi <Gabor.Csardi@unil.ch> UNIL DGM
Reply to: