[Rcpp-devel] Rcpp advocacy question -- build environment

Dirk Eddelbuettel edd at debian.org
Tue Nov 27 15:41:00 CET 2012


On 27 November 2012 at 13:57, Silkworth,David J. wrote:
| I have been attempting to collect enough material together to create a wiki
| book of sorts for the Windows user that is likely new to C++ as well.  I would
| be pleased to receive progress notes/problems off-list as you venture here.

That would be of help.
  
| Rtools is pretty good, but there are things happening that are quite foreign to
| most ?plug and play? Windows users.  The small bits of making sure you have
| assigned a correct PATH and, of course, assuring that R is installed in a
| directory without spaces are critical and will be overlooked by some no matter
| how you post it.  This stuff is much easier than getting a class of
| Wndows-oriented students going with a Linux installation.
|  
| There is a new wrinkle with the advent of 64-bit architectures when it comes to
| multi-arch package builds.  I think you would be best to stick to inline code
| development for just an introduction within a single semester sampler course of
| statistical programming.

Well -- I don't do much Windows, but when I do for testing, multiarch "just
works". And so it does for CRAN.

So please be careful do not spread misconceptions.  Yes, Windows use may well
be regarded as more difficult to setup and offering a fair amount of rope to
hang oneself --- but it works once setup right.

And as JJ explained in an earlier email, it works easier from RStudio as
additional checks are done by that environment too.  This also renders your
inline recommendation inadmissible because package building a) works under
Windows as is and b) is essentially automated.

We do understand that a lot of people are fond of inline. We love it too. But
there is **no reason whatsoever** to stop at inline, or recommend against
package building, or recommend against multiarch.  If things are broken on
your system, or if you have trouble with this, it is at your end. Do not
extrapolate.

Dirk

-- 
Dirk Eddelbuettel | edd at debian.org | http://dirk.eddelbuettel.com  


More information about the Rcpp-devel mailing list