[R-pkg-devel] packaging biosig for R

[Duncan Murdoch]

On 16/03/2020 3:00 a.m., Alois Schlögl wrote:
> 
> Thanks for your response. For *nix (Linux, MacOSX), this can be
> achieved. For windows, currently the only way to compile libbiosig is
> through mingw-cross-compiler environment (MXE) [1]  or in cygwin.
> Currently, libbiosig does not compile with with VC++for various reasons,
> and libbiosig has some hard dependencies (libiconv, suitesparse, dcmtk),
> (there are dependencies on libb64, tinyxml), I'm wondering what the best
> approach would be.
> 
> - Would it be ok, if libbiosig.dll is distributed in binary form ?

I think the answer to that is that it would not be allowed, but I'm not 
part of CRAN.

> 
> - Or would you accept a package that currently runs only on
> MacOSX/Homebrew and Linux but not on Windows ?

There are some packages like that on CRAN, but it is certainly 
discouraged.  Bioconductor has its own rules; you might want to ask 
there, given the subject matter of your package.

Duncan Murdoch


> 
> 
> Alois Schlögl
> 
> [1] https://github.com/schloegl/mxe
> 
> 
> 
> Am 3/16/20 um 2:08 AM schrieb Duncan Murdoch:
>> I suspect the question in this case depends on the availability of
>> libbiosig.  If CRAN test machines don't have that and its source isn't
>> included in the R package, then it will fail on initial install.  CRAN
>> doesn't have a lot of resources to install difficult libraries; I have
>> no idea if that applies in this case.
>>
>> I believe the most robust way to handle this sort of package is to
>> have a configure script that looks for the lib on the installing
>> machine, and uses that copy if found, otherwise compiles it from source.
>>
>> Duncan Murdoch
>>
>>
>>
>> On 15/03/2020 3:41 p.m., Jeff Newmiller wrote:
>>> I am just a lurker (not representing CRAN) but I am having a hard
>>> time understanding your question.
>>>
>>> Binary packages are a convenience for users, not a method for
>>> submitting packages. When you have an R package accepted it is
>>> accepted in source format.
>>> If it doesn't exclude support for Windows or MacOSX then it will (in
>>> time) be compiled into a binary form for distribution in addition to
>>> being distributed is source form.
>>>
>>> As the maintainer, your responsibility is merely to confirm that your
>>> source package is properly configured to be built in binary form
>>> before you submit it to CRAN. This is normally accomplished by
>>> successfully building it as binary in a testing environment. There
>>> are various guides out there that can be helpful in accomplishing
>>> this, e.g. [1].
>>>
>>> [1] https://kbroman.org/pkg_primer/pages/cran.htm
>>>
>>>
>>> On March 15, 2020 1:07:41 AM PDT, "Alois Schlögl"
>>> <alois.schloegl using gmail.com> wrote:
>>>>
>>>> Dear R packagers,
>>>>
>>>>
>>>> the Biosig project [1] supports reading of about 50 different data
>>>> format [2]. Recently, a language binding to R was added, because a user
>>>> of Biosig asked for it.
>>>>
>>>>
>>>> I've read the policy [3], and it seems the biosig would qualify as
>>>> binary package. The underlying library (libbiosig) can be installed
>>>>
>>>> - on linux from source, or through debian/ubuntu package
>>>>
>>>> - on MacOSX through Homebrew.
>>>>
>>>> - for Windows I'm using MXE mingw-cross-compiler environment to build
>>>> libbiosig.dll
>>>>
>>>>
>>>> Would it be feasible to provide a package of biosig on cran ? What need
>>>> to be considered ?
>>>>
>>>>
>>>> Kind regards
>>>>
>>>>      Alois
>>>>
>>>>
>>>>
>>>> [1] http://biosig.sourceforge.net/download.html
>>>>
>>>> [2] http://pub.ist.ac.at/~schloegl/biosig/TESTED
>>>>
>>>> [3] https://cran.r-project.org/web/packages/policies.html
>>>>
>>>> ______________________________________________
>>>> R-package-devel using r-project.org mailing list
>>>> https://stat.ethz.ch/mailman/listinfo/r-package-devel
>>>
>>