[Rd] How to ship R scripts with R packages ?

[Duncan Murdoch]

On 03/09/2009 7:47 AM, Romain Francois wrote:
> On 09/02/2009 01:13 PM, Duncan Murdoch wrote:
>> Romain Francois wrote:
>>> Hello,
>>>
>>> (This is a remix of this previous thread:
>>> https://stat.ethz.ch/pipermail/r-devel/2009-August/054264.html , but
>>> with a concrete example)
>>>
>>> I am developing some packages that contain scripts (for Rscript) and
>>> would like to know what is the best/recommended way to ship these
>>> scripts.
>>>
>>> An example is the "ant" package (R capable version of apache ant, see
>>> http://tr.im/xHLs). The package has a ant.R script in the exec
>>> directory and here is how I use it :
>>>
>>> $ `Rscript -e "cat( system.file( 'exec', 'ant.R', package = 'ant' ) ) " `
>>>
>>> ... not so pretty. Meanwhile, "Writing R extensions" talks about the
>>> exec directory, but it is not clear if I can use it this way or how.
>>>
>> I think exec is a reasonable place to put the script. If you don't want
>> to type that long command above, why not put it in a small function in
>> the package? E.g. define
>>
>> antR <- function() cat( system.file( 'exec', 'ant.R', package = 'ant' ) )
>>
>> then you can run it with
>>
>> Rscript -e "ant::antR()"
>>
>> Duncan Murdoch
> 
> I've implemented a minimal version of this in package "exec". For example.
> 
> $ Rscript -e "exec::rscript( 'ant', 'ant.R')"
> 
> Romain
> 
> btw, exec is maintained at r-forge :
> 
> $ svn checkout svn://svn.r-forge.r-project.org/svnroot/highlight/pkg/exec

That's a nice, simple function.  I might add an argument

dir = "exec"

rather than hard-coding that directory choice; others might keep their 
scripts somewhere else, or want to run one of the test scripts, etc.

Duncan Murdoch