[R-SIG-Mac] RQuantLib missing binaries

[Simon Urbanek]

On Jun 10, 2010, at 2:15 AM, Prof Brian Ripley wrote:

> On Wed, 9 Jun 2010, Kyle Matoba wrote:
> 
>> Just wondering if there was any plans to get this up and running again:
>> 
>> http://www.r-project.org/nosvn/R.check/r-release-macosx-ix86/RQuantLib-00install.html
>> 
>> If not, I will compile it myself.
> 
> You may not find it easy.  The issue is that RQuantLib depends on QuantLib, in fact on a recent version of QuantLib (later than the one in the current Fedora distribution, for example).  QuantLib is a large C++ suite of programs, and I've failed to compile it on Linux in the past, and when I have succeeded the package failed its own tests.
> 
>> I don't see anything in the changelog to indicate that this is
>> deliberate.  Could whomever is compiling for macs look into this?
> 
> It is done by Simon Urbanek's autobuiilder.  I don't see anything which indicates that it is not deliberate ....
> 
> If you look at the packages which are not being built by the Mac autobuiilder you will see three main reasons why:
> 
> (a) the package fails its tests, e.g. lme4
> (b) the package depends on other packages which are not available on the build machine, usually from BioC or OmegaHat.
> (c) the package depends on external software.
> 
> RQuantLib is in category (c), and very few such packages are being built (not even Simon's own packages GDD and proj4).  

If a package depends on external software it is built only if a) there is some demand for it and b) the dependencies can be built into self-contained static libraries with reasonably moderate effort. 

Now, since a) has been satisfied by this e-mail I was able to compile QuantLib and it is now available on the CRAN machine, but as Brian pointed out RQuantLib doesn't even pass its own checks, so despite our efforts there will be no binary. If you still want to build it despite it failing its own checks, you can get the QuantLib binary from
http://r.research.att.com/libs/
(beware, it's huge) and put the "boost" directory (containing the headers) from the Boost source distribution in /usr/local/include and you'll be able to compile RQuantLib for yourself.

Cheers,
Simon



> If you look at the CRAN test logs, the only platform on which RQuantLib is being installed is 32-bit Windows (not even Debian on which it is developed) -- and that is because the author supplied a pre-compiled Windows version of QuantLib.
> 
> I suspect you seriously underestimate the work which goes into providing R binary packages.  I know (I used to do it on Windows and still contribute there) just how thankless (literally and metaphorically) it is.
> 
> -- 
> 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
> 
> _______________________________________________
> R-SIG-Mac mailing list
> R-SIG-Mac at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
> 
>