On 16/06/2016 9:56 AM, Michael Cysouw wrote:
I really like the possibility of projects like docopt <http://docopt.org>, providing
a bash interface to R functionality. IMHO this is a great way to allow non-R-users to
include R functionality in their workflows. There is a docopt package for R which is not
yet perfect for my liking, but already goes a long way.
<https://cran.r-project.org/web/packages/docopt/index.html>
I have been experimenting a bit with including docopt scripts into my R
packages, and I would like to ask for some feedback how to handle the
distribution of such bash-interfaces with R packages.
Ideally, such bash-interfaces like docopt should be distributed together with
the R packages, and the “exec” folder seems destined for that goal. Simply
adding the package docopt as a requirement in DESCRIPTION also assures that
everything just works after installing the package!
The main remaining problem is the following: How do non-R-users actually find
the bash-executable? The location is rather hidden away :-). Is there a way to
link the path to the exec folder to PATH on installing a package? Or is there
an easy way to allow people to make an alias in their .bashrc?
Just put a description of what to do into your vignette.
From within R this is not that difficult, using something based on
“find.package”:
`file.path(find.package(package_name), "exec", docopt_file_name)`
However, I am more thinking about an option for R CMD INSTALL to link an
executable to PATH. Would that be sensible? How should that be done?
I don't think packages should be messing with users' PATHs. Document
what is necessary, even provide functions in your package to do the
work, but don't do it automatically. I think you were suggesting
something optional, but I really see no point in that: it is no more
work for you to write a function to do the work than it would be for you
to handle a request from INSTALL to do the work, but it is a lot more
work for us.
Duncan Murdoch
Further note that to prevent clogging the PATH with executable, I would propose
that package developers only include a single bash-interface to a package,
named identically to the package, with multiple functions available as
(sub)commands, so a bash user could simply access a function “bla” from package
“foo” as
$ foo bla -OPTIONS [ARGUMENTS]
thanks for any feedback
michael
______________________________________________
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
______________________________________________
R-package-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-package-devel
