[ADMB Users] ADMB as a backbone of an R package - how difficult is this?

[H. Skaug]

Hi,

I recommend calling the .exe using system() from R in the same way as
does glmmADMB, rather than using .dll's.
With the .exe approach you have all the ADMB output files (.par .std
.cor) residing in a directory which can
be used for diagnostic checking in case of problems, and in
communication with the ADMB user list.
At least you should use the .exe approach as a start to get your whole
system working, and get familiar
with ADMB. On the surface, communication via text files does not look
elegant, but you
then stay closer to the ADMB community, which is good in case your
engine is ADMB.

Hans


On Mon, Jul 9, 2012 at 9:54 PM, Philipp Doebler
<philipp.doebler at googlemail.com> wrote:
> Hello,
>
> I am in the process of working on an R-package
> (http://r-forge.r-project.org/projects/mada/). A central model in the
> package is currently fitted using optim(), which is acceptable, but
> since I plan to extend the model further, optim gets less and less
> attractive, especially with simulations in mind. After reading about
> ADMB I toyed around with it and realized that the executables produced
> can be used in an R-package.
>
> I studied the structure of the glmmADMB package and realized that I
> probably cannot just use the C++ code produced by ADMB and include it
> in my R-package, but I will have to produce executables for each
> platform, including .dll files. I am aware that producing dll-files is
> not straightforward as of now, but nevertheless I am eager to try. I
> am not fluent in any higher programming language, so using ADMB feels
> like a convenient route. So my questions are:
>
> 1) How much time/effort does one have to put into producing .dll
> files, files for mac os x, linux 32 bit and other non-native systems,
> granted that the ADMB model itself is already written and working on
> one platform? (I am working on a 64bit linux and I am far from an
> experienced software developer.).
>
> 2) Are there any posts on this list or other howtos for dll production?
>
> 3) I believe I am only using a tiny fraction of ADMB's functions, so
> are there ways to isolate this portion of functions and just include
> them as c++ header files in an R-package? Are there objections to this
> with respect to licenses? (my package is currently under GPL)
>
>
> With best wishes, Philipp
> _______________________________________________
> Users mailing list
> Users at admb-project.org
> http://lists.admb-project.org/mailman/listinfo/users