[R] Installing RQuantLib on Win 7 64 Bit

[Prof Brian Ripley]

On Fri, 26 Nov 2010, Mike Marchywka wrote:

> ----------------------------------------
>> Date: Thu, 25 Nov 2010 21:22:54 -0600
>> To: santosh.srinivas at gmail.com
>> From: edd at debian.org
>> CC: r-help at r-project.org
>> Subject: Re: [R] Installing RQuantLib on Win 7 64 Bit
>>
>>
>> On 26 November 2010 at 07:05, Santosh Srinivas wrote:
>> | Hello Group,
>> |
>> | I am trying out RQuantLib on a 64bit Win 7 machine. But running into
>> | installation errors
>>
>> The error message is about as clear as it can get:
>>
>> | install.packages("RQuantLib")
>> |
>> | Warning in install.packages("RQuantLib") :
>> | argument 'lib' is missing: using
>> | 'C:\Users\Tester\Documents/R/win64-library/2.11'
>> | Warning: unable to access index for repository
>> | http://www.stats.ox.ac.uk/pub/RWin/bin/windows64/contrib/2.11

Which indicated that this was R 2.11.x, a version of R for x64 Windows 
that is no longer supported and which used a different toolchain from 
current R builds.  No more binary packages will be produced for that 
platform, and its users were asked to update to pre-2.12.0 two months 
ago.

>> | Warning message:
>> | In getDependencies(pkgs, dependencies, available, lib) :
>> | package ‘RQuantLib’ is not available
>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>>
>> There is your answer: there simply is no binary package.
>>
>> I need a win64 development box (which I currently do not have) to build
>> QuantLib as a win64 library so that CRAN and R-Forge can turn the RQuantLib
>> source into a binary for you.

It is quite possible that could be done by cross-compiling: most of 
the external software used by win-builder and R-forge has been 
cross-compiled on (i686 or x86_64) Linux, including complex projects 
with C++ interfaces such as GDAL and SYMPHONY.  (C++ interfaces are 
much less portable than C interfaces so you do need carefully to match 
the cross-compiler to the toolchain used to compile the R package.)

> How bad are things getting? LOL. Seriously though, can anyone such as OP
> or I download source, build, install, and contrib it back?

Well, not 'anyone' if you mean contribute back to CRAN or R-forge. 
System maintainers do care about security, and to accept binary 
software will want verifiable credentials.  But 'anyone' *can* set up 
a CRAN-style repository and distribute binary packages from there 
(provided they follow the license conditions).  And additional 
repositories can be made known to an R installation by editing a text 
file (see ?setRepositories), or used directly by the 'contriburl' 
argument to install.packages() etc.

> Up until recently I had been building all the packages from source
> but one failed to install and now I just use install.package as was
> attempted here. I've been building on cygwin 1.7.

Which is of course a different OS, hosted on Windows.

-- 
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