Skip to content

Rscript on Windows

9 messages · Duncan Murdoch, Gabor Grothendieck, elijah wright +1 more

#
I mentioned this twice already and no one answered;however, I am mentioning
this a third time since its a serious deficiency.   The Rscript facility
that is upcoming in R is useful but on Windows one will often be relegated
to having two files: a batch file and an R file unless the -x switch
is implemented
to allow them to be combined.  This is not a problem on UNIX which supports
#! but on Windows we need -x.  Every other common scripting language including
perl, python and ruby supports -x for this purpose.

(The -x flag would start R processing at the first line that begins with #! so
that prior lines could be Windows batch commands allowing the same file
to be used as a batch file and an R file.)

Note that there is a bug in Windows which means that if you simply associate
.R to running R then the result cannot be redirected.  There is a bug
fix available
for this but I think we need to be able to run out of the box for something this
common.
On 1/29/07, Gabor Grothendieck <ggrothendieck at gmail.com> wrote:
#
On 2/16/2007 9:35 AM, Gabor Grothendieck wrote:
I agree this would be a reasonable addition, but I wouldn't class it as 
a serious deficiency, and I don't plan to work on it myself.

If you want to put together patches to the trunk code and docs to 
implement this I'll review them and possibly commit them.  If you don't 
see this as a high enough priority to do that, then I'd suggest doing 
what I do:  don't use the CMD.EXE shell.  There are a number of 
Unix-like shells available in Windows (Cygwin, MSYS, etc.) that can 
handle the #! syntax just fine.  Or just use two files, as you describe 
below.

Duncan Murdoch

 > The Rscript facility
#
I think its best if core mods are done by the core group while others
focus on work that can be done external to the core.

Thus, what I have done is to enhance the batchfiles distribution with
3 new batchfiles: Rscript.bat, #Rscript.bat and runR.bat which will be
part of the
next distribution of batchfiles but can be obtained now, if desired, from the
batchfiles svn (with the caveat that they require R 2.5.0).  The batchfiles
home page is here:

   http://code.google.com/p/batchfiles

The source tab on that page gets you to the svn and the links on the right
include links to the NEWS and README files which describe the additions,
a link to info on the Windows bug that I mentioned and two perl links that
describe how this all works in perl which may be a helpful analogous
situation.
.
On 2/17/07, Duncan Murdoch <murdoch at stats.uwo.ca> wrote:
#
On 2/17/2007 7:31 AM, Gabor Grothendieck wrote:
Fair enough, but then you also have to accept that the core group is 
going to set the priorities.  As far as I know *nobody* in the core 
group uses the CMD.EXE shell regularly, so changes to accommodate its 
limitations are going to get low priority.

Duncan Murdoch
#
Surely R has higher standards than that.  How about quality and
completeness of implementation?

Every other major scripting language has implemented this for good reason
and its a glaring omission.
On 2/17/07, Duncan Murdoch <murdoch at stats.uwo.ca> wrote:
#
Gabor, can we get a URL from you to a patch that implements this 
functionality?

Thanks!

--elijah
#
On Sat, 2007-02-17 at 09:31 -0500, Gabor Grothendieck wrote:
I think you are forgetting that R is an open source project, and is
reliant on the generous efforts of the R community, and in particular
the Core development team, for any work done on it.

I disagree with your statement about core mods being best done by the
core group - we all benefit when anyone, core or otherwise, contributes
to R. Duncan has already offered to review a submitted patch and
therefore commit some of his time to improving this feature - and this
is how it should work for those features that are of lower priority to
the core team.

However, that is your opinion and you are free to contribute directly to
R or not or contribute in some other way (as R-help subscribers know you
do to their benefit). But Core developers have that same right, and I'm
sure there are numerous other things in R that they might consider to be
incompletely implemented, in need of improvement or just plain missing
and therefore more deserving of their attention. Otherwise the SVN logs
wouldn't be quite so active...

G
#
Just because its open source does not mean everyone should do everything.
I suspect I have more expertise in Windows batch than the core developers
and also suspect they have more knowledge of the core than I so its a
good division of labor if I provide the batch files and they add -x
since it takes me
less time to produce batch files and they less time to add -x.

Furthermore its possible to develop something for the core and then have
it rejected and while hopefully this won't happen or if does it happens for
good reason, if one works on an external package then there is no chance
the work will be wasted since you have control over it.

In general the idea of having external packages has seemed to work well
and allows parallel development in a maximal way so the idea of having
the core work on the core and others work externally has been successful.

While no one has to provide Rscript or -x or anything else that applies to me
too and I didn't have to develop the external supporting Windows-specific
software in batchfiles or make it available to yet it is now available for you
and others to use with R 2.5.0 (via svn and more formally when I release it
to CRAN probably when R 2.5.0 is released).

Also I think that the success of R in the community is such that the core
developers do have some responsibility to the community at large beyond
their own needs.much as a business which when it gets to a certain size
and prominence has certain responsibilities to society beyond its own
purposes and some reasonable compromise between their own needs
and obvious requirements to complete certain work or do it to a certain
level of quality needs to be taken account of.
On 2/17/07, Gavin Simpson <gavin.simpson at ucl.ac.uk> wrote:
#
On 2/17/2007 10:57 AM, Gabor Grothendieck wrote:
[ deletions ]
I'd agree with this, as long as you don't limit it to the core 
developers.  We all have some responsibility to the community at large 
beyond our own needs.  But I think we also all have the right to decide 
how to prioritize those responsibilities.

much as a business which when it gets to a certain size
Sounds like you should contact Microsoft, and try to get them to take 
their responsibilities seriously, and support #! syntax in CMD.EXE. 
They've got a lot more resources than the R core group has.

Duncan Murdoch