R devel - Dec 2007 - Fortran 90 and Windows

I have been revising some FORTRAN 77 routines in R packages I have 
previously submitted.  Since R is now using gfortan I experimented with 
some Fortran 90 code (array intrinsics primarily).  So far the code is 
still in F77 fixed format, in files suffixed .f (not .f90), but 
incorporates some F90 constructs.  It has worked fine in linux/R.  I 
tried to follow the thread of previous discussions on this subject, but 
as fast as things are changing, it was somewhat dated, and not too 
definitive.

I don't so my own Windows versions of packages, but rather rely on CRAN 
to do the conversions, and I don't want to send them code that won't 
work on Windows.  Does anybody have extensive experience on what 
elements of F90 can be used in R packages for Windows?  If R CMD SHLIB 
is happy with it, is it likely to work?

Thanks, Dave
-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
David W. Roberts                                     office 406-994-4548
Professor and Head                                      FAX 406-994-3190
Department of Ecology                         email droberts at montana.edu
Montana State University
Bozeman, MT 59717-3460

Dave Roberts wrote:
> I have been revising some FORTRAN 77 routines in R packages I have 
> previously submitted.  Since R is now using gfortan I experimented with 
> some Fortran 90 code (array intrinsics primarily).  So far the code is 
> still in F77 fixed format, in files suffixed .f (not .f90), but 
> incorporates some F90 constructs.  It has worked fine in linux/R.  I 
> tried to follow the thread of previous discussions on this subject, but 
> as fast as things are changing, it was somewhat dated, and not too 
> definitive.
>
> I don't so my own Windows versions of packages, but rather rely on CRAN
> to do the conversions, and I don't want to send them code that
won't
> work on Windows.  Does anybody have extensive experience on what 
> elements of F90 can be used in R packages for Windows?  If R CMD SHLIB 
> is happy with it, is it likely to work?
>
> Thanks, Dave
>   
The Windows toolkit is also GCC, hence gfortran for new enough R. The
thing to worry about is the "other" category of machines. Either
oldish/specialized Unixes which ship with their own toolchains and
optimized libraries, or commercial compilers like the ones from Intel.
Brian Ripley and/or http://cran.r-project.org/doc/manuals/R-exts.html
can fill you in on the details. See also Brian's reply to John Fox on
December 9 ("I wouldn't let that deter you.").

-- 
   O__  ---- Peter Dalgaard             ?ster Farimagsgade 5, Entr.B
  c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
 (*) \(*) -- University of Copenhagen   Denmark          Ph:  (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)                  FAX: (+45) 35327907

On Mon, 17 Dec 2007, Dave Roberts wrote:
> I have been revising some FORTRAN 77 routines in R packages I have
> previously submitted.  Since R is now using gfortan I experimented with
> some Fortran 90 code (array intrinsics primarily).  So far the code is
> still in F77 fixed format, in files suffixed .f (not .f90), but
> incorporates some F90 constructs.  It has worked fine in linux/R.  I
> tried to follow the thread of previous discussions on this subject, but
> as fast as things are changing, it was somewhat dated, and not too
> definitive.
'Writing R Extensions' is both up-to-date and definitive, so please read
the primary documentation.  In particular, we have strengthened the 
comments about not using Fortran I/O, which can cause severe problems when 
using the GUI version of R under Windows (although we now have a 
workaround).
> I don't so my own Windows versions of packages, but rather rely on CRAN
> to do the conversions, and I don't want to send them code that
won't
> work on Windows.  Does anybody have extensive experience on what
> elements of F90 can be used in R packages for Windows?  If R CMD SHLIB
> is happy with it, is it likely to work?
Since Windows' builds nowadays use gcc 4.2.1, if it works on your 
unspecified version of gfortran on Linux it will very likely work on 
Windows.  And you can check that for yourself via 
win-builder.r-project.org.

However, using F9x code suffixed .f is not portable, and there are 
plenty of other systems where it will fail, some of which would accept 
.f90 or .f95 extensions.

See also my reply to John Fox:

https://stat.ethz.ch/pipermail/r-devel/2007-December/047695.html

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595